Axi-cel vs Liso-cel Analysis

Open libraries

Show code

library(gtsummary)
library(dplyr)
library(tidyverse)
library(haven)
library(survival)
library(survminer)
library(lubridate)
library(readxl)
library(janitor)
library(rms)
library(ggplot2)
library(tidycmprsk)
library(ggsurvfit)
library(caret)
library(MatchIt)
library(forester)
library(hrbrthemes)
library(WeightIt)
library(MatchIt)
library(cobalt)
library(survey)
library(ggsignif)

library(marginaleffects)
library(swimplot)
library(RColorBrewer)

Read in data

Read in tracking spreadsheet

Show code

db = read_excel(path = "Dataset 1.27.24.xlsx")

db <- db %>%
  filter(CAR.T.Product.Type == "Breyanzi" | CAR.T.Product.Type == "Yescarta") %>% # Include both Breyanzi and Yescarta
  filter(Acalabrutinib ==0) %>% # Remove patients treated on FH10418 RG1006269 acalabrutinib + yescarta
  filter(NKTR == 0) %>% # Remove patients treated on FH10802 RG1122036 NKTR-255 + breyanzi
  filter(Anakinra == 0) %>% # Remove patients treated on FH10373 RG1006866 Breyanzi + Anakinra
  filter(Disease != "BL" & Disease != "FL") %>% # Remove BL and FL
  filter(Infused == 1) %>% # Only include patients who received a CAR-T product
  mutate(
    Best.CR = ifelse(Best.response=="CR","1","0"),
    Best.PR = ifelse(Best.response=="PR","1","0"),
    Best.PD = ifelse(Best.response=="PD","1","0"),
    CRS.2.to.4 = ifelse(CRS.grade >=2,"1","0"),
    ICANS.2.to.4 = ifelse(ICANS.grade >=2, "1","0"),
    CRS.grade = factor(CRS.grade, levels = c("0","1","2","3","4")),
    ICANS.grade = factor(ICANS.grade, levels = c("0","1","2","3","4")),
    DOB = ymd(DOB),
    Leuka.date = ymd(Leuka.date),
    LD.date = ymd(LD.date),
    Infusion.Date = ymd(Infusion.Date),
    CRS.date = ymd(CRS.date),
    ICANS.date = ymd(ICANS.date),
    Initial.response.date = ymd(Initial.response.date),
    Date.of.relapse = ymd(Date.of.relapse),
    Date.of.death.or.last.contact = ymd(Date.of.death.or.last.contact),
    CRS.day = as.numeric(ifelse(CRS.grade !=0, CRS.date - Infusion.Date,"")),
    ICANS.day = as.numeric(ifelse(ICANS.grade !=0, ICANS.date - Infusion.Date,"")),
    CRS.ICANS.diff = as.numeric(ifelse(CRS.grade != 0 & ICANS.grade !=0, ICANS.date - CRS.date,"")),
    Days.to.DLC2 = Date.of.death.or.last.contact - Infusion.Date,
    Days.to.relapse2 = Date.of.relapse - Infusion.Date,
    Days.to.response = Initial.response.date - Infusion.Date,
    Vein.to.vein.time = Infusion.Date - Leuka.date,
    Bulky = ifelse(Largest.lesion >=5, "1","0")
    #Best.response = factor(Best.response, levels = c("CR","PR","SD","Mixed","PD","")),
    #Initial.response = factor(Initial.response, levels = c("CR","PR","SD","Mixed","PD",""))
  )

db$Best.CR <- as.factor(db$Best.CR)
db$Best.PR <- as.factor(db$Best.PR)
db$Best.PD <- as.factor(db$Best.PD)

Read in demographic data and join with master dataset

Show code

demographics = read_excel(path = "/Users/aportugu/Desktop/CD19 project/EMR data/demographics/2023.12.13.demographics.xlsx")

dataset <- 
  left_join(
    x = db,
    y = select(demographics, "ArrivalDate", "uwid"),
    by = "uwid"
  )

dataset <- dataset %>%
  mutate(
    ArrivalDate = ymd(ArrivalDate),
    Arrival.to.vein.time = Infusion.Date - ArrivalDate
  )

Read in product order data and join with master dataset

Show code

Order = read_excel(path = "/Users/aportugu/Desktop/CD19 project/EMR data/Order date/Order date.xlsx")

dataset <- 
  left_join(
    x = dataset,
    y = select(Order, "Order.date", "uwid"),
    by = "uwid"
  )

dataset <- dataset %>%
  mutate(
    Order.date = ymd(Order.date),
    Order.to.vein.time = Infusion.Date - Order.date
  )

Read in admissions data and join with master dataset

Show code

admissions = read_excel(path = "/Users/aportugu/Desktop/CD19 project/EMR data/admissions/2023.12.13.Admissions.xlsx")

inpatient.duration.df <- admissions %>% # Data frame for all admissions between days 0 and 30
  filter(admitDay %in% (0:30) & PatientClass == "Inpatient") %>%
  group_by(uwid) %>%
  summarise(inpatient.duration = sum(days), admissions = n())

elective.admission.df <- admissions %>% # Data frame for elective admissions only
  filter(admitDay == 0 & PatientClass == "Inpatient") %>% # Inpatient admission on day 0
  group_by(uwid) %>%
  summarise(elective.duration = sum(days)) %>%
  mutate(elective.admission = 1)

dataset <- # Join all admission data
  left_join(
    x = dataset,
    y = inpatient.duration.df,
    by = "uwid",
  )

dataset <- # Join elective admission data
  left_join(
    x = dataset,
    y = elective.admission.df,
    by = "uwid"
  )

dataset <- dataset %>%
  replace_na(list(inpatient.duration = 0, admissions = 0, elective.duration = 0, elective.admission = 0)) # Get rid of NA values

Read in CBC data

Show code

CBC = read_excel(path = "/Users/aportugu/Desktop/CD19 project/EMR data/labs/2023.12.13.hems.xlsx")

CBC$date = ymd(CBC$date)

CBC.df <- CBC %>% # Data frame for CBCs between days 0 and 30
  filter(anc != "" & day >=0 & day <=30) %>%
  group_by(uwid) %>%
  summarise(CBC.count = n(), WBC.min = min(wbc, na.rm=TRUE) / 1000, Plt.min = min(TotPlateletCnt, na.rm=TRUE) / 1000, Hgb.min = min(HgbAmt, na.rm=TRUE), ANC.min = min(anc, na.rm=TRUE) / 1000, ALC.min = min(lym, na.rm=TRUE) / 1000 )

CBC.30.df <- CBC %>% # Data frame for CBCs between days 0 and 30
  filter(anc != "" & day >=25 & day <=30) %>%
  group_by(uwid) %>%
  slice_max(day, with_ties = FALSE) %>%
  summarise(CBC.count = n(), WBC.30 = max(wbc, na.rm=TRUE) / 1000, Plt.30 = max(TotPlateletCnt, na.rm=TRUE) / 1000, Hgb.30 = max(HgbAmt, na.rm=TRUE), ANC.30 = max(anc, na.rm=TRUE) / 1000, ALC.30 = min(lym, na.rm=TRUE) / 1000 )
  
WBC.nadir.df <- CBC %>%
  filter(wbc != "" & day >=0 & day <=30) %>%
  group_by(uwid) %>%
  slice_min(wbc, with_ties = FALSE) %>%
  transmute(WBC_day_nadir = day)%>%
  select(WBC_day_nadir, uwid)

dataset <- # Join CBC data
  left_join(
    x = dataset,
    y = CBC.df,
    by = "uwid"
  )

dataset <- # Join ANC day nadir
  left_join(
    x = dataset,
    y = WBC.nadir.df,
    by = "uwid"
  )

dataset <- # Join ANC day nadir
  left_join(
    x = dataset,
    y = CBC.30.df,
    by = "uwid"
  )

neutropenic.df <- CBC %>%
  filter(anc != "" & anc <500 & day >=0 & day <=30) %>%
  group_by(uwid) %>%
  summarise(
    neutropenic.days = n()
  )

dataset <- # Join neutropenic days info
  left_join(
    x = dataset,
    y = neutropenic.df,
    by = "uwid"
  ) %>%
  mutate(
    neutropenic.days = ifelse(is.na(neutropenic.days),0,neutropenic.days)
  )


CBC.d0.df <- CBC %>% ## Data frame for day 0
  filter(day == 0) %>%
  group_by(uwid) %>%
  summarise(ANC.d0 = mean(anc, na.rm=TRUE) / 1000, ALC.d0 = mean (lym, na.rm=TRUE) / 1000)

dataset <-
  left_join(
    x = dataset,
    y = CBC.d0.df,
    by = "uwid"
  ) %>%
  mutate(
    ANC.d0 = ifelse(ANC.d0 == "NaN", NA, ANC.d0),
    ALC.d0 = ifelse(ALC.d0 == "NaN", NA, ANC.d0)
  )

Read in other lab data

Show code

labs = read_excel(path = "/Users/aportugu/Desktop/CD19 project/EMR data/labs/2023.12.13.labs.xlsx")

labs$date = ymd(labs$date)

labs.df <- labs %>% # Data frame for labs between days 0 and 30
  filter(day >=0 & day <=30) %>%
  group_by(uwid) %>%
  summarise(labs.count = n() , ALT.max = max(ALT_u_l, na.rm=TRUE), AST.max = max(AST_u_l, na.rm=TRUE), cre.min = min(CRE_mg_dl, na.rm=TRUE), cre.max = max(CRE_mg_dl, na.rm=TRUE), CRP.max = max(CRP_mg_l, na.rm=TRUE), ddimer.max = max(d_dimer, na.rm=TRUE), ferritin.max = max(ferritin_ng_ml, na.rm=TRUE), fibrinogen.min = min(fibrinogen_mg_dl, na.rm=TRUE), IL6.max = max(IL6_pg_ml, na.rm=TRUE), LDH.max = max(LD_u_l, na.rm=TRUE), LDH.min = min(LD_u_l, na.rm=TRUE), trop.max = max(troponin_ng_ml, na.rm=TRUE) ) 

labs.d3.df <- labs %>% ## Data frame for day 3
  filter(day == 3) %>%
  group_by(uwid) %>%
  summarise(ferritin.d3 = mean(ferritin_ng_ml, na.rm=TRUE), LDH.d3 = mean(LD_u_l, na.rm=TRUE), IL6.d3 = max(IL6_pg_ml, na.rm=TRUE),  CRP.d3 = max(CRP_mg_l, na.rm=TRUE))

labs.d0.df <- labs %>% ## Data frame for day 0
  filter(day == 0) %>%
  group_by(uwid) %>%
  summarise(ferritin.d0 = mean(ferritin_ng_ml, na.rm=TRUE), LDH.d0 = mean(LD_u_l, na.rm=TRUE), IL6.d0 = max(IL6_pg_ml, na.rm=TRUE) )

labs.df[sapply(labs.df, is.infinite)] <- NA ## Change any -inf values to NA
labs.d3.df[sapply(labs.d3.df, is.infinite)] <- NA ## Change any -inf values to NA
labs.d0.df[sapply(labs.d0.df, is.infinite)] <- NA ## Change any -inf values to NA

dataset <-
  left_join(
    x = dataset,
    y = labs.df,
    by = "uwid"
  )

dataset <- # Join ANC day nadir
  left_join(
    x = dataset,
    y = labs.d3.df,
    by = "uwid"
  )

dataset <- # Join ANC day nadir
  left_join(
    x = dataset,
    y = labs.d0.df,
    by = "uwid"
  )

Read in blood culture data

Show code

blood.culture = read_excel(path = "/Users/aportugu/Desktop/CD19 project/EMR data/id/2023.12.13.ID_ALL.xlsx")

blood.culture$date = ymd(blood.culture$date)

Pos.BCx <- blood.culture %>%
  select(uwid, bact, organism, date, day) %>%
  group_by(uwid) %>%
  filter(bact == "1" & day >=0) %>%
  distinct() %>%
  slice_min(day) %>%
  transmute(
    uwid,
    Pos.BCx = 1,
    BCx.day = day,
    BCx.date = date,
    BCx.organism = organism,
  )

Pos.CMV <- blood.culture %>%
  select(uwid, viral, organism, date, day, value) %>%
  group_by(uwid) %>%
  filter(organism == "CMV" & day >=0) %>%
  distinct() %>%
  summarise(CMV.first.day = min(day, na.rm = TRUE), CMV.last.day = max(day, na.rm = TRUE), CMV.peak.level = max(value, na.rm = TRUE)) %>%
  mutate(
    CMV = 1
  )

Pos.CMV[sapply(Pos.CMV, is.infinite)] <- NA ## Change any -inf values to NA

dataset <- # Join BCx data
  left_join(
    x = dataset,
    y = Pos.BCx,
    by = "uwid"
  )

dataset <- # Join CMV data
  left_join(
    x = dataset,
    y = Pos.CMV,
    by = "uwid"
  )

dataset <- dataset %>%
  mutate(
    Pos.BCx = ifelse( is.na(Pos.BCx) ,0,Pos.BCx),
    CMV = ifelse( is.na(CMV), 0, CMV ),
    CMV.peak.level = ifelse( is.na(CMV.peak.level), 0, CMV.peak.level)
  )

Determine the baseline pre-LD lab values

Show code

avg = 1

days = 14

CBC.with.LD.date <- # join the LD dates to the CBC dataframe
  left_join(
    x = CBC,
    y = dataset %>% select(uwid, LD.date),
    by = "uwid"
  )

## CBC
preLD.CBC.df <- CBC.with.LD.date %>%
  filter(LD.date >= date & LD.date - date <days & !is.na(wbc) ) %>%
  group_by(uwid) %>%
  slice_min(LD.date - date, n=avg) %>%
  summarise(WBC.preLD = mean(wbc, na.rm=TRUE) / 1000, Plt.preLD = mean(TotPlateletCnt, na.rm=TRUE) / 1000, Hgb.preLD = mean(HgbAmt, na.rm=TRUE))

dataset <- # join pre-LD CBC lab values
  left_join(
    x = dataset,
    y = preLD.CBC.df,
    by = "uwid"
  )

## Differential (ALC and ANC)
preLD.ALC.ANC.df <- CBC.with.LD.date %>%
  filter(LD.date >= date & LD.date - date <days & !is.na(anc) ) %>%
  group_by(uwid) %>%
  slice_min(LD.date - date, n=avg) %>%
  summarise(ANC.preLD = mean(anc, na.rm=TRUE) / 1000, ALC.preLD = mean (lym, na.rm=TRUE) / 1000)

dataset <- # join pre-LD CBC lab values
  left_join(
    x = dataset,
    y = preLD.ALC.ANC.df,
    by = "uwid"
  )



labs.with.LD.date <- # join the LD dates to the labs dataframe
  left_join(
    x = labs,
    y = dataset %>% select(uwid, LD.date),
    by = "uwid"
  )


## LDH
preLD.LDH.df <- labs.with.LD.date %>%
  filter(LD.date >= date & LD.date - date <days & !is.na(LD_u_l) ) %>% 
  group_by(uwid) %>%
  slice_min(LD.date - date, n=avg, na_rm = TRUE) %>%
  summarise(LDH.preLD = mean(LD_u_l, na.rm=TRUE))

dataset <- # join pre-LD lab values
  left_join(
    x = dataset,
    y = preLD.LDH.df,
    by = "uwid"
  )

dataset <- dataset %>%
  mutate(
    High.LDH = ifelse(LDH.preLD >=210,1,0),
    LDH.status = factor( case_when(
      LDH.preLD <210 ~ 0,
      LDH.preLD >=210 & LDH.preLD <=420 ~ 1,
      LDH.preLD >420 ~ 2), 
      levels = c(0,1,2) )
  )



## Renal function
preLD.Cr.df <- labs.with.LD.date %>%
  filter(LD.date >= date & LD.date - date <days & !is.na(CRE_mg_dl) ) %>%
  group_by(uwid) %>%
  slice_min(LD.date - date, n=avg, na_rm = TRUE) %>%
  summarise(cre.preLD = mean(CRE_mg_dl, na.rm=TRUE))

dataset <- # join pre-LD lab values
  left_join(
    x = dataset,
    y = preLD.Cr.df,
    by = "uwid"
  )


## Ferritin
preLD.ferritin.df <- labs.with.LD.date %>%
  filter(LD.date >= date & LD.date - date <days & !is.na(ferritin_ng_ml) ) %>%
  group_by(uwid) %>%
  slice_min(LD.date - date, n=avg, na_rm = TRUE) %>%
  summarise(ferritin.preLD = mean(ferritin_ng_ml, na.rm=TRUE))

dataset <- # join pre-LD lab values
  left_join(
    x = dataset,
    y = preLD.ferritin.df,
    by = "uwid"
  )


## CRP
preLD.CRP.df <- labs.with.LD.date %>%
  filter(LD.date >= date & LD.date - date <days & !is.na(CRP_mg_l) ) %>% 
  group_by(uwid) %>%
  slice_min(LD.date - date, n=avg, na_rm = TRUE) %>%
  summarise(CRP.preLD = mean(CRP_mg_l, na.rm=TRUE))

dataset <- # join pre-LD lab values
  left_join(
    x = dataset,
    y = preLD.CRP.df,
    by = "uwid"
  )

Read in BP data and determine pre-LD baseline

Show code

BP = read_excel(path = "/Users/aportugu/Desktop/CD19 project/EMR data/vs/2023.12.13.BP.xlsx")

BP$date = ymd(BP$date)

BP.post.df <- BP %>% # Data frame for labs between days 0 and 30
  filter(day >=0 & day <=30) %>%
  group_by(uwid) %>%
  summarise( BP.count = n() , SBP.max = max(SBP, na.rm=TRUE), DBP.max = max(DBP, na.rm=TRUE), SBP.min = min(SBP, na.rm=TRUE), DBP.min = min(DBP, na.rm=TRUE) ) 

dataset <- # post-infusion values
  left_join(
    x = dataset,
    y = BP.post.df,
    by = "uwid"
  )


BP.with.LD.date <- # join the LD dates to the BP dataframe
  left_join(
    x = BP,
    y = dataset %>% select(uwid, LD.date),
    by = "uwid"
  )

preLD.BP.df <- BP.with.LD.date %>%
  filter(LD.date >= date & LD.date - date <10) %>% ## Within 14 days prior to LD
  group_by(uwid) %>%
  slice_min(LD.date - date, n=3) %>%
  summarise(SBP.preLD = mean(SBP, na.rm=TRUE), DBP.preLD = mean(DBP, na.rm=TRUE) )

dataset <- # join pre-LD BP lab values
  left_join(
    x = dataset,
    y = preLD.BP.df,
    by = "uwid"
  ) %>%
  mutate(
    MAP.preLD = DBP.preLD + (SBP.preLD - DBP.preLD)/3,
    MAP.min = DBP.min + (SBP.min - DBP.min)/3
  )

Read in tocilizumab data

Show code

toci = read_excel(path = "/Users/aportugu/Desktop/CD19 project/EMR data/meds/2023.12.13.tocilizumab.xlsx")
toci$date = ymd(toci$date)

toci.df <- toci %>%
  select(uwid, date, day, Dose) %>%
  group_by(uwid) %>%
  filter(day >=0) %>%
  summarise(toci.first.day = min(day, na.rm=TRUE), toci.first.date = min(date, na.rm=TRUE), toci.last.day = max(day, na.rm=TRUE), toci.last.date = max(date, na.rm=TRUE), toci.doses = n())

dataset <- # join pre-LD CBC lab values
  left_join(
    x = dataset,
    y = toci.df,
    by = "uwid"
  ) %>%
  mutate(
    toci.doses = ifelse(is.na(toci.doses),0,toci.doses) # Set NA values to 0
  )

Read in dexamethasone data

Show code

dex = read_excel(path = "/Users/aportugu/Desktop/CD19 project/EMR data/meds/2023.12.13.dexamethasone.xlsx")
dex$date = ymd(dex$date)

dex.df <- dex %>%
  select(uwid, date, day, Dose, AdministrationRoute) %>%
  group_by(uwid) %>%
  filter(day >=0 & AdministrationRoute != "Oral" & AdministrationRoute != "PO" & AdministrationRoute != "EYE-Left" & AdministrationRoute != "EYE-Right") %>%
  summarise(dex.first.day = min(day, na.rm=TRUE), dex.first.date = min(date, na.rm=TRUE), dex.last.day = max(day, na.rm=TRUE), dex.last.date = max(date, na.rm=TRUE), dex.doses = n())

dataset <- # join pre-LD CBC lab values
  left_join(
    x = dataset,
    y = dex.df,
    by = "uwid"
  ) %>%
  mutate(
    dex.doses = ifelse(is.na(dex.doses),0,dex.doses) # Set NA values to 0
  )

Read in cefepime data

Show code

cefepime = read_excel(path = "/Users/aportugu/Desktop/CD19 project/EMR data/meds/2023.12.13.cefepime.xlsx")
cefepime$date = ymd(cefepime$date)

cefepime.df <- cefepime %>%
  select(uwid, date, day, Dose) %>%
  group_by(uwid) %>%
  filter(day >=0) %>%
  summarise(cefepime.first.day = min(day, na.rm=TRUE), cefepime.first.date = min(date, na.rm=TRUE), cefepime.last.day = max(day, na.rm=TRUE), cefepime.last.date = max(date, na.rm=TRUE), cefepime.doses = n())

dataset <- # join pre-LD CBC lab values
  left_join(
    x = dataset,
    y = cefepime.df,
    by = "uwid"
  ) %>%
  mutate(
    cefepime.doses = ifelse(is.na(cefepime.doses),0,cefepime.doses) # Set NA values to 0
  )

Read in vancomycin data

Show code

vancomycin = read_excel(path = "/Users/aportugu/Desktop/CD19 project/EMR data/meds/2023.12.13.vancomycin.xlsx")
vancomycin$date = ymd(vancomycin$date)

vancomycin.df <- vancomycin %>%
  select(uwid, date, day, Dose) %>%
  group_by(uwid) %>%
  filter(day >=0) %>%
  summarise(vancomycin.first.day = min(day, na.rm=TRUE), vancomycin.first.date = min(date, na.rm=TRUE), vancomycin.last.day = max(day, na.rm=TRUE), vancomycin.last.date = max(date, na.rm=TRUE), vancomycin.doses = n())

dataset <- # join pre-LD CBC lab values
  left_join(
    x = dataset,
    y = vancomycin.df,
    by = "uwid"
  ) %>%
  mutate(
    vancomycin.doses = ifelse(is.na(vancomycin.doses),0,vancomycin.doses) # Set NA values to 0
  )

Read in norepinephrine data

Show code

norepinephrine = read_excel(path = "/Users/aportugu/Desktop/CD19 project/EMR data/meds/2023.12.13.norepinephrine.xlsx")
norepinephrine$date = ymd(norepinephrine$date)

norepinephrine.df <- norepinephrine %>%
  select(uwid, date, day, Dose) %>%
  group_by(uwid) %>%
  filter(day >=0) %>%
  summarise(norepinephrine.first.day = min(day, na.rm=TRUE), norepinephrine.first.date = min(date, na.rm=TRUE), norepinephrine.last.day = max(day, na.rm=TRUE), norepinephrine.last.date = max(date, na.rm=TRUE), norepinephrine.doses = n())

dataset <- # join pre-LD CBC lab values
  left_join(
    x = dataset,
    y = norepinephrine.df,
    by = "uwid"
  ) %>%
  mutate(
    norepinephrine.doses = ifelse(is.na(norepinephrine.doses),0,norepinephrine.doses) # Set NA values to 0
  )

Read in temp data

Show code

temperature = read_excel(path = "/Users/aportugu/Desktop/CD19 project/EMR data/vs/2023.12.13.temperature.xlsx")
temperature$date = ymd(temperature$date)

fever.df <- temperature %>%
  select(uwid, date, day, temp_C) %>%
  group_by(uwid) %>%
  filter(day >=0 & temp_C >=38) %>%
  summarise(fever.first.day = min(day, na.rm=TRUE), fever.first.date = min(date, na.rm=TRUE), fever.last.day = max(day, na.rm=TRUE), fever.last.date = max(date, na.rm=TRUE), fever.days = n_distinct(day))

dataset <- # join pre-LD CBC lab values
  left_join(
    x = dataset,
    y = fever.df,
    by = "uwid"
  )

Read in ICE score data

Show code

ICE = read_excel(path = "/Users/aportugu/Desktop/CD19 project/EMR data/vs/2023.12.13.ICEscores.xlsx")
ICE$date = ymd(ICE$date)

ICE.df <- ICE %>%
  select(uwid, date, day, ICE.SCORE, ICE.ATTENTION, ICE.FOLLOW.COMMANDS, ICE.NAME.OBJECTS, ICE.ORIENTATION, ICE.WRITING) %>%
  group_by(uwid) %>%
  filter(day >=0) %>%
  summarise(ICE.min = min(ICE.SCORE, na.rm=TRUE), ICE.attention.min = min(ICE.ATTENTION, na.rm=TRUE), ICE.follow.commands.min = min(ICE.FOLLOW.COMMANDS, na.rm=TRUE), ICE.name.objects.min = min(ICE.NAME.OBJECTS, na.rm=TRUE), ICE.orientation.min = min(ICE.ORIENTATION, na.rm=TRUE), ICE.writing.min = min(ICE.WRITING, na.rm=TRUE))

dataset <- # join pre-LD CBC lab values
  left_join(
    x = dataset,
    y = ICE.df,
    by = "uwid"
  )

ICE.less.than.10.df <- ICE %>%
  select(uwid, date, day, ICE.SCORE, ICE.ATTENTION, ICE.FOLLOW.COMMANDS, ICE.NAME.OBJECTS, ICE.ORIENTATION, ICE.WRITING) %>%
  group_by(uwid) %>%
  filter(day >=0 & ICE.SCORE <10) %>%
  summarise(ICE.less.than.10 = n_distinct(day))

dataset <- # join pre-LD CBC lab values
  left_join(
    x = dataset,
    y = ICE.less.than.10.df,
    by = "uwid"
  )

All BP data points

Show code

BPdata = filter(BP, day >=-15 ) %>%
  select(day, SBP, DBP, time, uwid) %>%
  mutate(
    time.percent = sapply( strsplit(time,":"),
                           function(x) {
                             x <- as.numeric(x)
                             x[1]/24
                           }
    ),
    day = ifelse(day>=0, day + time.percent, day - time.percent),
    MAP = SBP/3 + DBP * 2/3
  )

BPdata <- # join pre-LD CBC lab values
  left_join(
    x = BPdata,
    y = dataset %>%
      select(CAR.T.Product.Type, uwid, Study.No, CRS.grade, ICANS.grade),
    by = "uwid"
  )

All ANC data points

Show code

CBCdata = filter(CBC, day >=-15 ) %>%
  select(day, ancx, lymx, HgbAmt, TotPlateletCnt, time, uwid) %>%
  mutate(
    time.percent = sapply( strsplit(time,":"),
                           function(x) {
                             x <- as.numeric(x)
                             x[1]/24
                           }
    ),
    day = ifelse(day>=0, day + time.percent, day - time.percent),
  )

CBCdata <-
  left_join(
    x = CBCdata,
    y = dataset %>%
      select(CAR.T.Product.Type, uwid, Study.No),
    by = "uwid"
  )

Propensity scores

Regression analysis for treatment allocation

Show code

variables <- c("CAR.T.Product.Type", "Age", "Second.line", "LDH.preLD", "ALC.preLD", "ASCT", "Largest.lesion", "tDLBCL", "ECOG.status", "Extranodal", "Bridging.response.category")

data1 <- dataset %>% 
  drop_na(LDH.preLD, ALC.preLD, Bridging.response.category, ECOG) %>%
  mutate(
    Bulky = ifelse(Largest.lesion >=5,1,0),
    ECOG.status = ifelse(ECOG >=2,1,0),
    ALC.low = ifelse(ALC.preLD <=1,1,0),
    CAR.T.Product.Type = recode(CAR.T.Product.Type, "Breyanzi" = 1, "Yescarta" = 0),
    Bridging.response.category = factor(recode(Bridging.response.category, "0" = "No bridging", "1" = "CR/PR", "2" = "SD/PD", "3"="Unknown response"), levels = c("No bridging","CR/PR","SD/PD", "Unknown response")),
  )

uv_tab <- tbl_uvregression(
    data1[c(variables)],
    method = glm,
    y = CAR.T.Product.Type,
    method.args = list(family = binomial),
    exponentiate = TRUE
  ) %>%
  sort_p()


mv_tab<-glm(CAR.T.Product.Type ~ Age + Second.line + LDH.preLD + ALC.preLD + ASCT + Largest.lesion + tDLBCL + ECOG.status + Extranodal + Bridging.response.category, data=data1, family=binomial) %>%
  tbl_regression(exponentiate=TRUE)

tbl_merge(list(uv_tab, mv_tab), tab_spanner = c("**Univariate**", "**Multivariable**"))

Characteristic	Univariate				Multivariable
	N	OR1	95% CI1	p-value	OR1	95% CI1	p-value
Age	148	1.06	1.03, 1.10	<0.001	1.06	1.03, 1.11	0.001
Bridging.response.category	148
No bridging		—	—		—	—
CR/PR		3.30	1.23, 8.98	0.017	2.64	0.83, 8.46	0.10
SD/PD		1.77	0.76, 4.09	0.2	2.84	0.96, 8.80	0.062
Unknown response		46,954,082	0.00, NA	>0.9	38,025,973	0.00, NA	>0.9
Second.line	148	2.48	0.84, 7.50	0.10	1.95	0.51, 7.74	0.3
ASCT	148	0.46	0.15, 1.24	0.15	0.50	0.14, 1.49	0.2
ALC.preLD	148	1.41	1.00, 2.41	0.2	1.24	0.96, 2.24	0.3
tDLBCL	148	1.56	0.68, 3.51	0.3	1.70	0.61, 4.66	0.3
LDH.preLD	148	1.00	1.00, 1.00	0.3	1.00	1.00, 1.00	0.3
Largest.lesion	148	0.95	0.85, 1.05	0.3	0.93	0.81, 1.06	0.3
ECOG.status	148	0.74	0.29, 1.78	0.5	0.74	0.23, 2.16	0.6
Extranodal	148	0.88	0.44, 1.76	0.7	1.11	0.47, 2.68	0.8
1 OR = Odds Ratio, CI = Confidence Interval

IPTW

Show code

IPTW.data <- dataset %>% 
  #drop_na(LDH.preLD, ALC.preLD, Bridging.response.category, ECOG) %>%
  mutate(
    Age.65 = ifelse(Age >=65,1,0),
    Male.sex = ifelse(Sex == "M",1,0),
    Bulky = ifelse(Largest.lesion >=5,1,0),
    ECOG.status = ifelse(ECOG >=2,1,0)
  )

## Use the WeightIt function to generate propensity scores and weights
W.out <- weightit(CAR.T.Product.Type ~ Age + Second.line + ASCT + Largest.lesion + tDLBCL + ECOG.status + Extranodal, 
                  data=IPTW.data, 
                  method = "glm", 
                  estimand = "ATE",
                  stabilize = TRUE)
summary = summary(W.out)

bal.tab(W.out, stats = c("m", "v"), thresholds = c(m=0.25), binary = "std")

Balance Measures
                   Type Diff.Adj     M.Threshold V.Ratio.Adj
prop.score     Distance   0.0848 Balanced, <0.25      1.2909
Age             Contin.  -0.1721 Balanced, <0.25      1.5656
Second.line      Binary   0.0487 Balanced, <0.25           .
ASCT             Binary   0.0734 Balanced, <0.25           .
Largest.lesion  Contin.  -0.0718 Balanced, <0.25      0.3979
tDLBCL           Binary  -0.0855 Balanced, <0.25           .
ECOG.status      Binary  -0.0856 Balanced, <0.25           .
Extranodal       Binary  -0.0411 Balanced, <0.25           .

Balance tally for mean differences
                    count
Balanced, <0.25         8
Not Balanced, >0.25     0

Variable with the greatest mean difference
 Variable Diff.Adj     M.Threshold
      Age  -0.1721 Balanced, <0.25

Effective sample sizes
           Breyanzi Yescarta
Unadjusted    50.      99.  
Adjusted      40.48    86.18

Show code

love.plot(W.out,
          binary = "std",
          line = TRUE, 
          abs = TRUE,
          thresholds = c(m=0.25),
          var.order = "unadjusted",
            var.names = c(
              Age = "Age",
              prop.score = "Distance",
              Male.sex = "Male sex",
              Second.line = "Second line",
              LDH.preLD = "LDH pre-LD",
              ALC.preLD = "ALC pre-LD",
              Bridging = "Bridging",
              ASCT = "Prior ASCT",
              Largest.lesion = "Largest lesion",
              Bridging.response.category = "Bridging response",
              tDLBCL = "tDLBCL vs other",
              ECOG.status = "ECOG\u22652"
          ))

Show code

IPTW.data$weights <- W.out$weights * (summary$effective.sample.size$Breyanzi[2] + summary$effective.sample.size$Yescarta[2])/sum(W.out$weights) # Attach normalized weights to a dataset

IPTW.data$ps <- W.out$ps

PSM

Show code

PSM.data <- dataset %>% 
  #drop_na(LDH.preLD, ALC.preLD, Bridging.response.category, ECOG) %>%
  mutate(
    Age.65 = ifelse(Age >=65,1,0),
    Male.sex = ifelse(Sex == "M",1,0),
    Bulky = ifelse(Largest.lesion >=5,1,0),
    ECOG.status = ifelse(ECOG >=2,1,0),
    Bridging.response.category = factor(recode(Bridging.response.category, "0" = "No bridging", "1" = "CR/PR", "2" = "SD/PD", "3"="Unknown response"), levels = c("No bridging","CR/PR","SD/PD", "Unknown response")),
    ALC.low = ifelse(ALC.preLD <=1,1,0),
    CAR.T.Product.Type = ifelse(CAR.T.Product.Type == "Breyanzi",1,0)
  )


m.out <- matchit(CAR.T.Product.Type ~ Age + Second.line + ASCT + Largest.lesion + tDLBCL + ECOG.status + Extranodal, data=PSM.data, method = "full", distance = "glm", link = "logit") # Full matching


bal.tab(m.out, stats = c("m", "v"), thresholds = c(m=0.25), binary = "std")

Balance Measures
                   Type Diff.Adj     M.Threshold V.Ratio.Adj
distance       Distance  -0.0027 Balanced, <0.25      0.9453
Age             Contin.  -0.0776 Balanced, <0.25      1.1420
Second.line      Binary   0.1300 Balanced, <0.25           .
ASCT             Binary  -0.0300 Balanced, <0.25           .
Largest.lesion  Contin.  -0.0545 Balanced, <0.25      2.5635
tDLBCL           Binary   0.0027 Balanced, <0.25           .
ECOG.status      Binary  -0.0591 Balanced, <0.25           .
Extranodal       Binary  -0.2025 Balanced, <0.25           .

Balance tally for mean differences
                    count
Balanced, <0.25         8
Not Balanced, >0.25     0

Variable with the greatest mean difference
   Variable Diff.Adj     M.Threshold
 Extranodal  -0.2025 Balanced, <0.25

Sample sizes
                     Control Treated
All                    99.        50
Matched (ESS)          26.27      50
Matched (Unweighted)   99.        50

Show code

m.out_NN <- matchit(CAR.T.Product.Type ~ Age + Second.line + ASCT + Largest.lesion + tDLBCL + ECOG.status + Extranodal, data=PSM.data, method = "nearest", distance = "glm", link = "logit") # Nearest neighbor matching for comparison


love.plot(m.out, stats = c("m"), abs = TRUE,
          binary = "std",
          drop.distance = FALSE, 
          weights = list(nn = m.out_NN),
          var.order = "unadjusted",
          line = TRUE, 
          thresholds = c(m=0.25, ks = 0.05),
          sample.names = c("Original", "Full Matching", "Nearest Neighbor Matching"),
          position = "bottom",
          var.names = c(
            prop.score = "Distance",
            Bridging = "Bridging",
            LDH.preLD = "LDH pre-LD",
            LDH.status = "LDH status",
            ALC.preLD = "ALC pre-LD",
            Age.62 = "Age \u226562",
            Male.sex = "Male sex",
            Second.line = "Second line",
            ASCT = "Prior ASCT",
            Largest.lesion = "Largest lesion",
            HCT.CI = "HCT-CI",
            ECOG.status = "ECOG \u22652",
            Bridging.response.category = "Bridging response",
            tDLBCL = "tDLBCL vs other"

          )
          )

Show code

# Create a dataset with PSM weights using full matching
m.data <- match.data(m.out) %>%
  mutate(
    CAR.T.Product.Type = ifelse(CAR.T.Product.Type == 1, "Breyanzi","Yescarta")
  )

m.data_NN <- match.data(m.out_NN)

Scatterplot of weights

Show code

ggplot(IPTW.data %>% mutate(Product = CAR.T.Product.Type, Product = recode(Product, "Breyanzi"="Liso-cel","Yescarta"="Axi-cel")), aes(x=ps, y = weights, color = Product)) + 
  geom_point(shape = 16) +
  theme_classic() + 
  labs(x = "Propensity score", y = "Weights") # Add a legend label

Show code

ggplot(m.data, aes(x=distance, y = weights, color = CAR.T.Product.Type)) + 
  geom_point(shape = 16) +
  theme_minimal() + 
  ggtitle("PSM") +
  labs(x = "Propensity score", y = "Weights") # Add a legend label

TABLES

Baseline characteristics

Show code

theme_gtsummary_compact()

dataset %>%
  transmute(
    #"Age \u226565 years" = ifelse(Age >=65,1,0),
    "Age (years)" = Age,
    "Male sex" = ifelse(Sex == "M", "Yes","No"),
    "ECOG PS" = factor( recode(ECOG, "0"="0-1","1"="0-1","2"="\u22652","3"="\u22652","4"="\u22652"), levels = c("0-1", "\u22652") ),
    "HCT-CI" = HCT.CI,
    "TRANSCEND eligible" = TRANSCEND,
    "Second line" = Second.line,
    "Prior ASCT" = ASCT,
    #"Received bridging" = Bridging,
    "Bridging response category" = factor(recode(Bridging.response.category, "0" = "No bridging", "1" = "CR/PR", "2" = "SD/PD", "3" = "Unknown response"), levels = c("No bridging","CR/PR","SD/PD", "Unknown response")),
    "Disease subtype" = Disease,
    "LDH" = LDH.preLD,
    "ALC" = ALC.preLD,
    #"Elevated LDH (\u2265210 U/L)" = ifelse(LDH.preLD >=210,"Yes","No"),
    #"LDH status" = factor(recode(LDH.status, "0" = "\u2264ULN", "1" = "[ULN; 2x ULN]", "2"=">2x ULN"), levels = c("\u2264ULN", "[ULN; 2x ULN]", ">2x ULN") ),
    "Largest lesion (cm)" = Largest.lesion,
    #"Bulky (\u22655 cm)" = ifelse(Bulky == 1, "Yes","No"),
    "Extranodal" = ifelse(Extranodal == 1, "Yes","No"),
    "Out-of-spec" = ifelse(EAP == 1,"Yes","No"),
    "Product" = factor(recode(CAR.T.Product.Type, "Breyanzi"= "Liso-cel", "Yescarta" = "Axi-cel"), levels = c("Liso-cel","Axi-cel"))
  ) %>%
  tbl_summary(
    missing = "ifany",
    #sort = all_categorical() ~ "frequency",
    type = c("HCT-CI" = "continuous", "ECOG" = "continuous"),
    #type = c("HCT-CI" = "continuous", "ECOG" = "continuous", "Max CRS grade" = "continuous", "Max ICANS grade" = "continuous", "Total inpatient duration (days)" = "continuous", "Days with fever" ~ "continuous"),
    statistic = list(
      all_continuous() ~ "{median} ({p25}, {p75})",
      all_categorical() ~ "{n} ({p}%)",
      "Age (years)" ~"{median} ({min} to {max})"
      #"Total inpatient duration (days)" ~ "{median} ({min} to {max})"
      #"HCT-CI" ~ "{median} ({min} to {max})"
    )
  ) %>%
  bold_labels

Characteristic	N = 1491
Age (years)	62 (26 to 83)
Male sex	93 (62%)
ECOG PS
0-1	120 (81%)
≥2	29 (19%)
HCT-CI	1.00 (1.00, 3.00)
TRANSCEND eligible	102 (68%)
Second line	15 (10%)
Prior ASCT	24 (16%)
Bridging response category
No bridging	88 (59%)
CR/PR	21 (14%)
SD/PD	36 (24%)
Unknown response	4 (2.7%)
Disease subtype
DLBCL	104 (70%)
HGBCL	7 (4.7%)
PMBCL	5 (3.4%)
TCHRLBCL	2 (1.3%)
tDLBCL	31 (21%)
LDH	186 (151, 292)
Unknown	1
ALC	0.60 (0.33, 0.99)
Largest lesion (cm)	3.2 (1.6, 5.5)
Extranodal	87 (58%)
Out-of-spec	13 (8.7%)
Product
Liso-cel	50 (34%)
Axi-cel	99 (66%)
1 Median (Range); n (%); Median (IQR)

Show code

dataset %>%
  transmute(
    #"Age \u226565 years" = ifelse(Age >=65,1,0),
    "Age (years)" = Age,
    "Male sex" = ifelse(Sex == "M", "Yes","No"),
    "ECOG PS" = factor( recode(ECOG, "0"="0-1","1"="0-1","2"="\u22652","3"="\u22652","4"="\u22652"), levels = c("0-1", "\u22652") ),
    "HCT-CI" = HCT.CI,
    "TRANSCEND eligible" = TRANSCEND,
    "Second line" = Second.line,
    "Prior ASCT" = ASCT,
    #"Received bridging" = Bridging,
    "Bridging response category" = factor(recode(Bridging.response.category, "0" = "No bridging", "1" = "CR/PR", "2" = "SD/PD", "3" = "Unknown response"), levels = c("No bridging","CR/PR","SD/PD", "Unknown response")),
    "Disease subtype" = Disease,
    "LDH" = LDH.preLD,
    "ALC" = ALC.preLD,
    #"Elevated LDH (\u2265210 U/L)" = ifelse(LDH.preLD >=210,"Yes","No"),
    #"LDH status" = factor(recode(LDH.status, "0" = "\u2264ULN", "1" = "[ULN; 2x ULN]", "2"=">2x ULN"), levels = c("\u2264ULN", "[ULN; 2x ULN]", ">2x ULN") ),
    "Largest lesion (cm)" = Largest.lesion,
    #"Bulky (\u22655 cm)" = ifelse(Bulky == 1, "Yes","No"),
    "Extranodal" = ifelse(Extranodal == 1, "Yes","No"),
    "Out-of-specification" = ifelse(EAP == 1,"Yes","No"),
    "Product" = factor(recode(CAR.T.Product.Type, "Breyanzi"= "Liso-cel", "Yescarta" = "Axi-cel"), levels = c("Liso-cel","Axi-cel"))
  ) %>%
  tbl_summary(
    by = "Product",
    missing = "ifany",
    #sort = all_categorical() ~ "frequency",
    type = c("HCT-CI" = "continuous", "ECOG" = "continuous"),
    statistic = list(
      all_continuous() ~ "{median} ({p25}, {p75})",
      all_categorical() ~ "{n} ({p}%)",
      "Age (years)" ~"{median} ({min} to {max})"
    )
  ) %>%
  modify_spanning_header(all_stat_cols() ~ "**CAR-T Treatment**") %>%
  modify_header(all_stat_cols() ~ "**{level}**<br>N = {n} ({style_percent(p)}%)") %>%
  bold_labels %>%
  add_p() %>%
  add_stat_label()

Characteristic	CAR-T Treatment		p-value1
	Liso-cel N = 50 (34%)	Axi-cel N = 99 (66%)
Age (years), Median (Range)	67 (45 to 83)	61 (26 to 79)	<0.001
Male sex, n (%)	27 (54%)	66 (67%)	0.13
ECOG PS, n (%)			0.4
0-1	42 (84%)	78 (79%)
≥2	8 (16%)	21 (21%)
HCT-CI, Median (IQR)	1.00 (1.00, 3.00)	1.00 (1.00, 3.00)	>0.9
TRANSCEND eligible, n (%)	31 (62%)	71 (72%)	0.2
Second line, n (%)	8 (16%)	7 (7.1%)	0.087
Prior ASCT, n (%)	5 (10%)	19 (19%)	0.15
Bridging response category, n (%)			0.002
No bridging	22 (44%)	66 (67%)
CR/PR	11 (22%)	10 (10%)
SD/PD	13 (26%)	23 (23%)
Unknown response	4 (8.0%)	0 (0%)
Disease subtype, n (%)			0.4
DLBCL	33 (66%)	71 (72%)
HGBCL	3 (6.0%)	4 (4.0%)
PMBCL	0 (0%)	5 (5.1%)
TCHRLBCL	1 (2.0%)	1 (1.0%)
tDLBCL	13 (26%)	18 (18%)
LDH, Median (IQR)	180 (151, 245)	206 (153, 308)	0.3
Unknown	0	1
ALC, Median (IQR)	0.74 (0.44, 1.01)	0.57 (0.31, 0.96)	0.081
Largest lesion (cm), Median (IQR)	2.7 (0.2, 4.4)	3.4 (2.1, 6.6)	0.067
Extranodal, n (%)	28 (56%)	59 (60%)	0.7
Out-of-specification, n (%)	11 (22%)	2 (2.0%)	<0.001
1 Wilcoxon rank sum test; Pearson’s Chi-squared test; Fisher’s exact test

IPTW weighted data table

Show code

theme_gtsummary_compact()
data1 = IPTW.data %>%
  mutate(
    "Age (years)" = Age,
    Male.sex = ifelse(Sex == "M", "Yes","No"),
    ECOG = factor( recode(ECOG, "0"="0-1","1"="0-1","2"="\u22652","3"="\u22652","4"="\u22652"), levels = c("0-1", "\u22652") ),
    Bridging = factor(recode(Bridging.response.category, "0"="No bridging","1"="CR/PR","2"="SD/PD","3"="Unknown response"), levels = c("No bridging","CR/PR","SD/PD", "Unknown response")),
    Product = factor(recode(CAR.T.Product.Type, "Breyanzi"= "Liso-cel", "Yescarta" = "Axi-cel"), levels = c("Liso-cel","Axi-cel")),
  )

  survey::svydesign(
    id = ~0,
    weights = ~weights,
    data = data1,
    fpc = NULL
  ) %>%
  tbl_svysummary(
    by = Product,
    missing = "ifany",
    statistic = list(
      all_continuous() ~ "{median} ({p25}, {p75})",
      all_categorical() ~ "{n} ({p}%)"),
    include = c(Age, Male.sex, ECOG, HCT.CI, TRANSCEND, Second.line, ASCT, Disease, LDH.preLD, ALC.preLD, Largest.lesion, Extranodal, Bridging, EAP),
    label = list(
      Male.sex ~ "Male sex",
      HCT.CI ~ "HCT-CI",
      Second.line ~ "Second line",
      LDH.preLD ~ "LDH",
      ALC.preLD ~ "ALC",
      Largest.lesion ~ "Largest lesion (cm)",
      Bridging ~ "Bridging response category"
    ),
  ) %>%
  add_p() %>%
    bold_labels()#%>%

Characteristic	Liso-cel, N = 411	Axi-cel, N = 861	p-value2
Age	61 (52, 69)	64 (52, 68)	0.7
Male sex	24 (59%)	59 (69%)	0.3
ECOG			0.7
0-1	31 (76%)	68 (79%)
≥2	10 (24%)	18 (21%)
HCT-CI	1.00 (1.00, 3.00)	1.00 (1.00, 3.00)	0.9
TRANSCEND	22 (55%)	63 (73%)	0.053
Second line	4 (8.7%)	9 (10%)	0.8
ASCT	5 (13%)	13 (16%)	0.7
Disease			0.7
DLBCL	27 (67%)	59 (69%)
HGBCL	2 (4.2%)	3 (3.2%)
PMBCL	0 (0%)	3 (3.6%)
TCHRLBCL	1 (1.8%)	1 (1.0%)
tDLBCL	11 (27%)	20 (23%)
LDH	180 (151, 266)	190 (147, 297)	0.8
Unknown	0	1
ALC	0.56 (0.38, 0.91)	0.54 (0.31, 0.97)	0.5
Largest lesion (cm)	3.3 (0.0, 5.2)	3.4 (2.1, 5.9)	0.5
Extranodal	25 (61%)	51 (59%)	0.8
Bridging response category			0.020
No bridging	17 (42%)	55 (64%)
CR/PR	8 (19%)	12 (14%)
SD/PD	12 (30%)	18 (21%)
Unknown response	3 (7.9%)	0 (0%)
EAP	9 (23%)	1 (1.6%)	<0.001
1 Median (IQR); n (%)
2 Wilcoxon rank-sum test for complex survey samples; chi-squared test with Rao & Scott’s second-order correction

Show code

  #sort_p()

Reasons ineligible for TRANSCEND

Show code

data <- dataset %>%
  distinct() %>%
  filter(TRANSCEND == 0 & !is.na(Rationale.category) ) %>%
  count(Rationale.category) %>%
  mutate(
    pct = prop.table(n) * 100
    )

data %>%
  ggplot() +
  aes(x = "", y = pct, fill = reorder(Rationale.category, -pct)) +
  geom_col(position = 'stack', width = 1) +
  coord_polar("y") +
  xlab(NULL) +
  ylab(NULL) +
  theme_void() +
  scale_fill_brewer(palette = "Pastel2") +
  geom_text(aes(x = 1.3, label = paste0(sprintf("%1.f", pct), "%", "\u000A", "n=", n)), position = position_stack(vjust = 0.5), size = 5) +
  theme(axis.text.x = element_blank(),
        axis.ticks = element_blank(),
        legend.text = element_text(size = 16),
        legend.title = element_text(size = 16)) +  # Set legend title font size and make it bold
  labs(fill = paste0("Primary reason (N = ", tally(filter(dataset, TRANSCEND==0))$n, ")" ) )

Time to product infusion

Show code

theme_gtsummary_compact()

dataset %>%
  #filter(Largest.lesion !=0) %>% #only include measurable disease
  filter(Infused == 1) %>%
  transmute(
    "Product" = factor(recode(CAR.T.Product.Type, "Breyanzi"= "Liso-cel", "Yescarta" = "Axi-cel"), levels = c("Liso-cel","Axi-cel")),
    "Vein-to-vein time (days)" = Vein.to.vein.time,
    "Arrival-to-infusion time (days)" = Arrival.to.vein.time,
    "Order-to-infusion time (days)" = Order.to.vein.time,
    #"Total inpatient duration (days)" = inpatient.duration,
  ) %>%
  tbl_summary(
    by = "Product",
    missing = "no",
    statistic = list(
      all_continuous() ~ "{median} ({p25}, {p75})",
      all_categorical() ~ "{n} ({p}%)"
    )
  ) %>%
  bold_labels %>%
  add_p() %>%
  add_stat_label()

Characteristic	Liso-cel, N = 50	Axi-cel, N = 99	p-value1
Vein-to-vein time (days), Median (IQR)	34 (32, 40)	27 (26, 30)	<0.001
Arrival-to-infusion time (days), Median (IQR)	46 (42, 56)	40 (35, 45)	<0.001
Order-to-infusion time (days), Median (IQR)	62 (44, 77)	34 (32, 41)	<0.001
1 Wilcoxon rank sum test

Pre-LD lab values

Show code

theme_gtsummary_compact()

dataset %>%
  filter(LD == 1) %>%
  transmute(
    "Pre-LD WBC (10\U00B3/\U00B5L)" = WBC.preLD,
    "Pre-LD Plt (10\U00B3/\U00B5L)" = Plt.preLD,
    "Pre-LD Hgb (g/dL)" = Hgb.preLD,
    "Pre-LD ANC (10\U00B3/\U00B5L)" = ANC.preLD,
    "ALC pre-LD (10\U00B3/\U00B5L)" = ALC.preLD,
    "Pre-LD ferritin (ng/mL)" = ferritin.preLD,
    "Pre-LD cre (mg/dL)" = cre.preLD,
    "LDH pre-LD (U/L)" = LDH.preLD,
    "Product" = CAR.T.Product.Type,
  ) %>%
  tbl_summary(
    by = "Product",
    missing = "no",
    #sort = all_categorical() ~ "frequency",
    statistic = list(
      all_continuous() ~ "{median} ({p25}, {p75})",
      all_categorical() ~ "{n} ({p}%)"
    )
  ) %>%
  bold_labels %>%
  add_p() %>%
  sort_p() %>%
  add_stat_label()

Characteristic	Breyanzi, N = 50	Yescarta, N = 99	p-value1
ALC pre-LD (10³/µL), Median (IQR)	0.74 (0.44, 1.01)	0.57 (0.31, 0.96)	0.081
Pre-LD Hgb (g/dL), Median (IQR)	11.30 (10.05, 12.32)	10.20 (9.60, 11.90)	0.10
LDH pre-LD (U/L), Median (IQR)	180 (151, 245)	206 (153, 308)	0.3
Pre-LD ferritin (ng/mL), Median (IQR)	266 (119, 672)	388 (174, 836)	0.4
Pre-LD cre (mg/dL), Median (IQR)	0.84 (0.66, 1.01)	0.87 (0.70, 1.04)	0.6
Pre-LD ANC (10³/µL), Median (IQR)	2.67 (2.06, 4.21)	3.00 (1.61, 4.97)	0.6
Pre-LD Plt (10³/µL), Median (IQR)	149 (114, 199)	160 (99, 203)	0.8
Pre-LD WBC (10³/µL), Median (IQR)	4.2 (3.4, 6.3)	4.7 (3.0, 6.6)	>0.9
1 Wilcoxon rank sum test

Admission characteristics

Show code

theme_gtsummary_compact()

dataset %>%
  filter(Infused == 1) %>%
  transmute(
    "Product" = CAR.T.Product.Type,
    "Vein-to-vein time (days)" = Vein.to.vein.time,
    "Order-to-vein time (days)" = Order.to.vein.time,
    "Arrival-to-vein time (days)" = Arrival.to.vein.time,
    "# of admissions" = admissions,
    '2+ admissions' = ifelse(admissions>=2, 1, 0),
    "Elective admission" = recode(elective.admission, "1" = "Yes", "0" = "No"),
    "Total inpatient duration (days)" = inpatient.duration,
    "Elective duration (days)" = elective.duration,
    "Non-elective duration (days)" = inpatient.duration - elective.duration,
  ) %>%
  tbl_summary(
    by = "Product",
    missing = "no",
    #sort = all_categorical() ~ "frequency",
    type = c("Elective duration (days)" ~ "continuous"),
    statistic = list(
      all_continuous() ~ "{median} ({p25}, {p75})",
      all_categorical() ~ "{n} ({p}%)",
      "Elective duration (days)" ~ "{median} ({min} to {max})",
      "Non-elective duration (days)" ~ "{median} ({min} to {max})",
      "Total inpatient duration (days)" ~ "{median} ({min} to {max})"
    )
  ) %>%
  bold_labels %>%
  add_p() %>%
  add_stat_label()

Characteristic	Breyanzi, N = 50	Yescarta, N = 99	p-value1
Vein-to-vein time (days), Median (IQR)	34 (32, 40)	27 (26, 30)	<0.001
Order-to-vein time (days), Median (IQR)	62 (44, 77)	34 (32, 41)	<0.001
Arrival-to-vein time (days), Median (IQR)	46 (42, 56)	40 (35, 45)	<0.001
# of admissions, n (%)			<0.001
0	19 (38%)	9 (9.1%)
1	27 (54%)	69 (70%)
2	2 (4.0%)	18 (18%)
3	2 (4.0%)	3 (3.0%)
2+ admissions, n (%)	4 (8.0%)	21 (21%)	0.042
Elective admission, n (%)	5 (10%)	82 (83%)	<0.001
Total inpatient duration (days), Median (Range)	5 (0 to 35)	14 (0 to 54)	<0.001
Elective duration (days), Median (Range)	0 (0 to 30)	12 (0 to 48)	<0.001
Non-elective duration (days), Median (Range)	2.5 (0.0 to 35.0)	0.0 (0.0 to 54.0)	0.006
1 Wilcoxon rank sum test; Fisher’s exact test; Pearson’s Chi-squared test

Vein-to-vein histogram

Show code

Median_breyanzi_vv = signif(median(as.numeric(filter(dataset, CAR.T.Product.Type == "Breyanzi")$Vein.to.vein.time)), 2)
Median_yescarta_vv = signif(median(as.numeric(filter(dataset, CAR.T.Product.Type == "Yescarta")$Vein.to.vein.time)), 2)

histogram_vv <- dataset %>%
  mutate(CAR.T.Product.Type = recode(CAR.T.Product.Type,"Breyanzi" = "Liso-cel","Yescarta" = "Axi-cel")) %>%
  ggplot( aes(x=as.numeric(Vein.to.vein.time), fill=CAR.T.Product.Type) )  +
    geom_histogram( color="#e9ecef", alpha=0.6, position = 'identity', binwidth=7) +
    scale_fill_manual(values = c("#69b3a2", "#404080"), guide = "none") +  # Remove the legend
    theme_classic() +
    labs(fill="") + 
  geom_vline(aes(xintercept = Median_yescarta_vv),
             linetype = "dashed", color = "#69b3a2", linewidth = 0.75)+ 
  geom_vline(aes(xintercept = Median_breyanzi_vv),
             linetype = "dashed", color = "#404080", linewidth = 0.75) +
  xlim(10, 70)+
  xlab("Vein-to-vein time (days)") + 
  ylab("Number of patients") + 
  annotate("text", label=Median_yescarta_vv, x = Median_yescarta_vv +3, y = 61, color = "#69b3a2") + 
  annotate("text", label=Median_breyanzi_vv, x = Median_breyanzi_vv +3, y = 61, color = "#404080") + 
  theme(legend.position = c(0.8, 0.8)) + 
  theme(
    axis.title = element_text(size = 16),
    axis.text = element_text(size = 16),
    legend.title = element_text(size = 16),
    legend.text = element_text(size = 14)
  ) + 
    annotate("text", x = 45, y = 55, label = sprintf("p <.001"), size = 5)
histogram_vv

Age histogram

Show code

Median_breyanzi_age = signif(median(as.numeric(filter(dataset, CAR.T.Product.Type == "Breyanzi")$Age)), 2)
Median_yescarta_age = signif(median(as.numeric(filter(dataset, CAR.T.Product.Type == "Yescarta")$Age)), 2)

histogram_age <- dataset %>%
  mutate(CAR.T.Product.Type = recode(CAR.T.Product.Type,"Breyanzi" = "Liso-cel","Yescarta" = "Axi-cel")) %>%
  ggplot( aes(x=as.numeric(Age), fill=CAR.T.Product.Type) )  +
    geom_histogram( color="#e9ecef", alpha=0.6, position = 'identity', binwidth=5) +
    scale_fill_manual(values=c("#69b3a2", "#404080")) +
    theme_classic() +
    labs(fill="") + 
  geom_vline(aes(xintercept = Median_yescarta_age),
             linetype = "dashed", color = "#69b3a2", linewidth = 0.75)+ 
  geom_vline(aes(xintercept = Median_breyanzi_age),
             linetype = "dashed", color = "#404080", linewidth = 0.75) +
  xlim(20, 90)+
  xlab("Age (years)") + 
  ylab("Number of patients") + 
  annotate("text", label=Median_yescarta_age, x = Median_yescarta_age +3, y = 22, color = "#69b3a2") + 
  annotate("text", label=Median_breyanzi_age, x = Median_breyanzi_age +3, y = 22, color = "#404080") + 
  theme(legend.position = c(0.87, 0.8)) + 
  theme(
    axis.title = element_text(size = 16),
    axis.text = element_text(size = 16),
    legend.title = element_text(size = 16),
    legend.text = element_text(size = 12)
  ) + 
    annotate("text", x = 50, y = 20, label = sprintf("p <.001"), size = 5)
histogram_age

Combined histogram

Show code

ggarrange(histogram_vv, histogram_age, ncol = 2, labels = c("A", "B"))

Table for response

Show code

theme_gtsummary_compact()

data1 <- dataset %>%
  mutate(
    CAR.T.Product.Type = recode(CAR.T.Product.Type, "Breyanzi" = "Liso-cel","Yescarta" = "Axi-cel")
  )

tbl1 <- data1 %>%
  filter(Infused == 1) %>%
  filter(Response.evaluable ==1) %>% #only include measurable disease
  transmute(
    "Product" = CAR.T.Product.Type,
    "Initial response" = Initial.response,
    "Best response" = Best.response,
    "Progressive disease" = recode(Best.PD,"0"="No","1"="Yes"),
    "ORR" = ifelse(Best.response == "CR" | Best.response == "PR",1,0)
  ) %>%
  tbl_summary(
    by = "Product",
    missing = "no",
    #sort = all_categorical() ~ "frequency",
    statistic = list(
      all_continuous() ~ "{median} ({p25}, {p75})",
      all_categorical() ~ "{n} ({p}%)"
    )
  ) %>%
  bold_labels %>%
  add_p() %>%
  add_stat_label()

tbl2 <- data1 %>%
  filter(Infused == 1) %>%
  transmute(
    "Product" = CAR.T.Product.Type,
    "Initial response" = Initial.response,
    "Best response" = Best.response,
    "Progressive disease" = recode(Best.PD,"0"="No","1"="Yes"),
    "ORR" = ifelse(Best.response == "CR" | Best.response == "PR",1,0)
  ) %>%
  tbl_summary(
    by = "Product",
    missing = "no",
    #sort = all_categorical() ~ "frequency",
    statistic = list(
      all_continuous() ~ "{median} ({p25}, {p75})",
      all_categorical() ~ "{n} ({p}%)"
    )
  ) %>%
  bold_labels %>%
  add_p() %>%
  add_stat_label()

tbl_merge(
  tbls = list(tbl1, tbl2),                          # combine
  tab_spanner = c("**Response evaluable**", "**All patients**")) # set header names

Characteristic	Response evaluable			All patients
	Axi-cel, N = 95	Liso-cel, N = 41	p-value1	Axi-cel, N = 99	Liso-cel, N = 50	p-value1
Initial response, n (%)			0.7			>0.9
CR	39 (41%)	13 (32%)		42 (42%)	22 (44%)
PD	21 (22%)	10 (24%)		22 (22%)	10 (20%)
PR	33 (35%)	17 (41%)		33 (33%)	17 (34%)
SD	2 (2.1%)	1 (2.4%)		2 (2.0%)	1 (2.0%)
Best response, n (%)			>0.9			0.7
CR	45 (47%)	20 (49%)		48 (48%)	29 (58%)
PD	21 (22%)	10 (24%)		22 (22%)	10 (20%)
PR	28 (29%)	11 (27%)		28 (28%)	11 (22%)
SD	1 (1.1%)	0 (0%)		1 (1.0%)	0 (0%)
Progressive disease, n (%)	21 (22%)	10 (24%)	0.8	22 (22%)	10 (20%)	0.8
ORR, n (%)	73 (77%)	31 (76%)	0.9	76 (77%)	40 (80%)	0.7
1 Fisher’s exact test; Pearson’s Chi-squared test

IPTW weighted response table

Show code

theme_gtsummary_compact()
data1 = IPTW.data %>%
  transmute(
    "Product" = recode(CAR.T.Product.Type, "Breyanzi"="Liso-cel","Yescarta"="Axi-cel"),
    "Initial response" = Initial.response,
    "Best response" = Best.response,
    "Progressive disease" = as.factor( recode(Best.PD,"0"="No","1"="Yes") ),
    "ORR" = ifelse(Best.response == "CR" | Best.response == "PR",1,0),
    weights,
    Response.evaluable
  )


tbl1<-  survey::svydesign(
    id = ~0,
    weights = ~weights,
    data = filter(data1, Response.evaluable==1), # Only include measurable disease
    fpc = NULL
  ) %>%
  tbl_svysummary(
    by = Product,
    missing = "ifany",
    statistic = list(
      all_continuous() ~ "{median} ({p25}, {p75})",
      all_categorical() ~ "{n} ({p}%)"),
    include = c("Initial response", "Best response", "Progressive disease", "ORR")
  ) %>%
  bold_labels() %>%
  add_p()

tbl2<-  survey::svydesign(
    id = ~0,
    weights = ~weights,
    data = data1, # Include response evaluable and all patients
    fpc = NULL
  ) %>%
  tbl_svysummary(
    by = Product,
    missing = "ifany",
    statistic = list(
      all_continuous() ~ "{median} ({p25}, {p75})",
      all_categorical() ~ "{n} ({p}%)"),
    include = c("Initial response", "Best response", "Progressive disease", "ORR")
  ) %>%
  bold_labels() %>%
  add_p()
  
  
tbl_merge(
  tbls = list(tbl1, tbl2),                          # combine
  tab_spanner = c("**Response evaluable**", "**All patients**")) # set header names

Characteristic	Response evaluable			All patients
	Axi-cel, N = 821	Liso-cel, N = 341	p-value2	Axi-cel, N = 861	Liso-cel, N = 411	p-value2
Initial response			0.7			0.8
CR	36 (44%)	12 (34%)		38 (44%)	18 (45%)
PD	18 (22%)	7 (21%)		20 (24%)	7 (17%)
PR	26 (32%)	14 (43%)		26 (31%)	14 (36%)
SD	1 (1.8%)	1 (2.8%)		1 (1.7%)	1 (2.4%)
Best response			>0.9			0.6
CR	41 (50%)	18 (52%)		43 (50%)	24 (60%)
PD	18 (22%)	7 (21%)		20 (24%)	7 (17%)
PR	22 (27%)	9 (27%)		22 (26%)	9 (23%)
SD	1 (1.0%)	0 (0%)		1 (1.0%)	0 (0%)
Progressive disease	18 (22%)	7 (21%)	0.8	20 (24%)	7 (17%)	0.4
ORR	63 (77%)	27 (79%)	0.7	65 (75%)	34 (83%)	0.3
1 n (%)
2 chi-squared test with Rao & Scott’s second-order correction

Table for toxicity

Show code

theme_gtsummary_compact()

dataset %>%
  filter(Infused == 1) %>%
  transmute(
    "Product" = CAR.T.Product.Type,
    #"Max CRS grade" = CRS.grade,
    "CRS any grade" = ifelse( CRS.grade != 0, 1, 0),
    "CRS grade 3-4" = ifelse( CRS.grade ==3 | CRS.grade == 4, 1, 0),
    "Time to CRS (days)" = CRS.day,
    #"Max ICANS grade" = ICANS.grade,
    "ICANS any grade" = ifelse( ICANS.grade != 0,1,0),
    "ICANS grade 3-4" = ifelse( ICANS.grade ==3 | ICANS.grade == 4,1,0),
    "Days with fever" = fever.days,
    "Fever 1st day" = fever.first.day,
    "Time to ICANS (days)" = ICANS.day,
    "Toci doses" = toci.doses,
    "Received toci" = ifelse(toci.doses !=0,1,0),
    "Dex doses" = dex.doses,
    "Received dex" = ifelse(dex.doses !=0,1,0),
    "ICE score min" = ICE.min
  ) %>%
  tbl_summary(
    by = "Product",
    missing = "no",
    #sort = all_categorical() ~ "frequency",
    statistic = list(
      all_continuous() ~ "{median} ({p25}, {p75})",
      all_categorical() ~ "{n} ({p}%)"
      #"Time to ICANS (days)" ~ "{median} ({min}, {max})"
    ),
    type = list("Toci doses" ~ "continuous", "ICE score min" ~ "continuous", "Dex doses" ~ "continuous", "Time to ICANS (days)" ~ "continuous")
  ) %>%
  bold_labels %>%
  add_p() %>%
  add_stat_label()

Characteristic	Breyanzi, N = 50	Yescarta, N = 99	p-value1
CRS any grade, n (%)	32 (64%)	88 (89%)	<0.001
CRS grade 3-4, n (%)	1 (2.0%)	8 (8.1%)	0.3
Time to CRS (days), Median (IQR)	4.00 (2.00, 6.00)	4.00 (2.00, 6.00)	>0.9
ICANS any grade, n (%)	14 (28%)	52 (53%)	0.004
ICANS grade 3-4, n (%)	9 (18%)	18 (18%)	>0.9
Days with fever, Median (IQR)	3 (2, 4)	5 (3, 7)	<0.001
Fever 1st day, Median (IQR)	3.00 (1.50, 5.00)	3.00 (1.00, 5.00)	0.7
Time to ICANS (days), Median (IQR)	7.0 (7.0, 8.8)	7.0 (5.0, 9.3)	0.4
Toci doses, Median (IQR)	0.00 (0.00, 0.00)	1.00 (0.00, 1.00)	<0.001
Received toci, n (%)	12 (24%)	53 (54%)	<0.001
Dex doses, Median (IQR)	0 (0, 5)	2 (0, 13)	0.001
Received dex, n (%)	16 (32%)	65 (66%)	<0.001
ICE score min, Median (IQR)	10.0 (6.0, 10.0)	8.0 (3.0, 10.0)	0.013
1 Pearson’s Chi-squared test; Fisher’s exact test; Wilcoxon rank sum test

IPTW weighted toxicity table

Show code

theme_gtsummary_compact()
data1 = IPTW.data %>%
  transmute(
    "Product" = recode(CAR.T.Product.Type, "Breyanzi"="Liso-cel","Yescarta"="Axi-cel"),
    #"Max CRS grade" = CRS.grade,
    "CRS any grade" = ifelse( CRS.grade != 0, 1, 0),
    "CRS grade 3-4" = ifelse( CRS.grade ==3 | CRS.grade == 4, 1, 0),
    "Time to CRS (days)" = CRS.day,
    #"Max ICANS grade" = ICANS.grade,
    "ICANS any grade" = ifelse( ICANS.grade != 0,1,0),
    "ICANS grade 3-4" = ifelse( ICANS.grade ==3 | ICANS.grade == 4,1,0),
    "Days with fever" = fever.days,
    "Fever 1st day" = fever.first.day,
    "Time to ICANS (days)" = ICANS.day,
    "Toci doses" = as.numeric(toci.doses),
    "Received toci" = ifelse(toci.doses !=0,1,0),
    "Dex doses" = as.numeric(dex.doses),
    "Received dex" = ifelse(dex.doses !=0,1,0),
    "ICE score min" = ICE.min,
    weights
  )


survey::svydesign(
    id = ~0,
    weights = ~weights,
    data = data1,
    fpc = NULL
  ) %>%
  tbl_svysummary(
    by = Product,
    missing = "no",
    statistic = list(
      all_continuous() ~ "{median} ({p25}, {p75})",
      all_categorical() ~ "{n} ({p}%)"),
    include = c("CRS any grade","CRS grade 3-4","ICANS any grade","ICANS grade 3-4")
  ) %>%
  bold_labels %>%
  add_p()

Characteristic	Axi-cel, N = 861	Liso-cel, N = 411	p-value2
CRS any grade	73 (85%)	27 (67%)	0.040
CRS grade 3-4	7 (7.8%)	1 (3.0%)	0.3
ICANS any grade	43 (50%)	15 (36%)	0.2
ICANS grade 3-4	14 (17%)	10 (25%)	0.3
1 n (%)
2 chi-squared test with Rao & Scott’s second-order correction

Table for ID complications

Show code

theme_gtsummary_compact()

dataset %>%
  transmute(
    "Product" = CAR.T.Product.Type,
    "Positive BCx" = Pos.BCx,
    "Positive CMV" = CMV,
    "CMV level detectable" = ifelse(CMV.peak.level >0, 1, 0),
    "CMV duration" = CMV.last.day - CMV.first.day,
    "Cefepime doses" = cefepime.doses,
    "Received cefepime" = ifelse(cefepime.doses >=1,1,0),
    "Vancomycin doses" = vancomycin.doses
  ) %>%
  tbl_summary(
    by = "Product",
    missing = "no",
    #sort = all_categorical() ~ "frequency",
    statistic = list(
      all_continuous() ~ "{median} ({p25}, {p75})",
      all_categorical() ~ "{n} ({p}%)"
    )
  ) %>%
  bold_labels %>%
  add_p() %>%
  add_stat_label() %>%
  sort_p()

Characteristic	Breyanzi, N = 50	Yescarta, N = 99	p-value1
Cefepime doses, Median (IQR)	0 (0, 8)	9 (0, 16)	<0.001
Received cefepime, n (%)	21 (42%)	68 (69%)	0.002
Vancomycin doses, Median (IQR)	0.0 (0.0, 0.0)	0.0 (0.0, 3.0)	0.3
Positive BCx, n (%)	2 (4.0%)	8 (8.1%)	0.5
CMV level detectable, n (%)	6 (12%)	9 (9.1%)	0.6
Positive CMV, n (%)	15 (30%)	26 (26%)	0.6
CMV duration, Median (IQR)	19 (4, 26)	14 (7, 34)	0.8
1 Fisher’s exact test; Pearson’s Chi-squared test; Wilcoxon rank sum test

Table for post-infusion cytopenias

Show code

theme_gtsummary_compact()

data1 <- dataset %>%
  mutate(
    CAR.T.Product.Type = recode(CAR.T.Product.Type,"Breyanzi"="Liso-cel","Yescarta"="Axi-cel")
  )

data1 %>%
  filter(Infused == 1) %>%
  transmute(
    "Product" = CAR.T.Product.Type,
    "WBC min (10\U00B3/\U00B5L)" = WBC.min,
    "WBC nadir day" = WBC_day_nadir,
    "ANC min (10\U00B3/\U00B5L)" = ANC.min,
    "ANC <500 (cells/\U00B5L)" = ifelse(ANC.min <0.5, "Yes","No"),
    "Hgb min (g/dL)" = Hgb.min,
    "Plt min (10\U00B3/\U00B5L)" = Plt.min,
    "Neutropenic time (days)" = neutropenic.days
  ) %>%
  tbl_summary(
    by = "Product",
    missing = "no",
    #sort = all_categorical() ~ "frequency",
    statistic = list(
      all_continuous() ~ "{median} ({p25}, {p75})",
      all_categorical() ~ "{n} ({p}%)"
    )
  ) %>%
  bold_labels %>%
  add_p() %>%
  add_stat_label() %>%
  sort_p()

Characteristic	Axi-cel, N = 99	Liso-cel, N = 50	p-value1
ANC min (10³/µL), Median (IQR)	0.04 (0.00, 0.17)	0.26 (0.10, 0.51)	<0.001
WBC min (10³/µL), Median (IQR)	0.25 (0.14, 0.48)	0.65 (0.37, 1.38)	<0.001
Neutropenic time (days), Median (IQR)	6.0 (3.0, 10.0)	2.0 (0.3, 5.0)	<0.001
Hgb min (g/dL), Median (IQR)	8.20 (7.75, 8.85)	8.80 (8.25, 10.10)	0.001
ANC <500 (cells/µL), n (%)	92 (93%)	37 (74%)	0.001
Plt min (10³/µL), Median (IQR)	36 (18, 67)	82 (32, 109)	0.002
WBC nadir day, Median (IQR)	4.0 (3.0, 6.0)	5.0 (3.0, 6.8)	0.4
1 Wilcoxon rank sum test; Pearson’s Chi-squared test

Table for day 30 counts

Show code

theme_gtsummary_compact()

data1 <- dataset %>%
  mutate(
    CAR.T.Product.Type = recode(CAR.T.Product.Type,"Breyanzi"="Liso-cel","Yescarta"="Axi-cel")
  )

data1 %>%
  transmute(
    "Product" = CAR.T.Product.Type,
    "WBC d30 (10\U00B3/\U00B5L)" = WBC.30,
    "ANC d30 (10\U00B3/\U00B5L)" = ANC.30,
    "ALC d30 (10\U00B3/\U00B5L)" = ALC.30,
    "ANC <500 (cells/\U00B5L)" = ifelse(ANC.30 <0.5, "Yes","No"),
    "Hgb d30 (g/dL)" = Hgb.30,
    "Plt d30 (10\U00B3/\U00B5L)" = Plt.30,
  ) %>%
  tbl_summary(
    by = "Product",
    missing = "no",
    #sort = all_categorical() ~ "frequency",
    statistic = list(
      all_continuous() ~ "{median} ({p25}, {p75})",
      all_categorical() ~ "{n} ({p}%)"
    )
  ) %>%
  bold_labels %>%
  add_p() %>%
  add_stat_label() %>%
  sort_p()

Characteristic	Axi-cel, N = 99	Liso-cel, N = 50	p-value1
Plt d30 (10³/µL), Median (IQR)	63 (33, 117)	102 (48, 145)	0.049
ANC d30 (10³/µL), Median (IQR)	1.38 (0.78, 2.31)	1.72 (1.15, 2.95)	0.14
WBC d30 (10³/µL), Median (IQR)	2.59 (1.53, 3.73)	2.90 (2.12, 4.07)	0.4
ANC <500 (cells/µL), n (%)	14 (15%)	5 (10%)	0.5
ALC d30 (10³/µL), Median (IQR)	0.47 (0.24, 0.95)	0.48 (0.33, 0.63)	0.5
Hgb d30 (g/dL), Median (IQR)	10.40 (9.40, 11.55)	10.50 (9.95, 11.20)	0.6
1 Wilcoxon rank sum test; Pearson’s Chi-squared test

Table for post-infusion inflammatory markers and other labs

Show code

theme_gtsummary_compact()

dataset %>%
  mutate(
    CAR.T.Product.Type = recode(CAR.T.Product.Type,"Breyanzi"="Liso-cel","Yescarta"="Axi-cel")
  ) %>%
  transmute(
    "Product" = CAR.T.Product.Type,
    "ALT max (U/L)" = ALT.max,
    "AST max (U/L)" = AST.max,
    "CRP max (mg/L)" = CRP.max,
    "D-dimer max (mg/L)" = ddimer.max,
    "Ferritin max (ng/mL)" = ferritin.max,
    "IL6 max (pg/mL)" = IL6.max,
    "LDH max (U/L)" = LDH.max
  ) %>%
  tbl_summary(
    by = "Product",
    missing = "no",
    #sort = all_categorical() ~ "frequency",
    statistic = list(
      all_continuous() ~ "{median} ({p25}, {p75})",
      all_categorical() ~ "{n} ({p}%)"
    )
  ) %>%
  bold_labels %>%
  add_p() %>%
  add_stat_label() %>%
  sort_p()

Characteristic	Axi-cel, N = 99	Liso-cel, N = 50	p-value1
CRP max (mg/L), Median (IQR)	118 (57, 180)	63 (9, 114)	<0.001
IL6 max (pg/mL), Median (IQR)	207 (58, 1,165)	55 (15, 481)	0.001
ALT max (U/L), Median (IQR)	53 (36, 82)	30 (22, 67)	0.004
D-dimer max (mg/L), Median (IQR)	2.5 (1.5, 3.9)	1.4 (0.7, 3.3)	0.007
Ferritin max (ng/mL), Median (IQR)	1,217 (488, 2,520)	783 (206, 1,709)	0.017
AST max (U/L), Median (IQR)	36 (27, 60)	28 (22, 49)	0.073
LDH max (U/L), Median (IQR)	250 (189, 355)	215 (176, 346)	0.2
1 Wilcoxon rank sum test

Table for post-infusion inflammatory markers and other labs, stratified by product and CRS

Show code

theme_gtsummary_compact()

dataset %>%
  mutate(
    CAR.T.Product.Type = recode(CAR.T.Product.Type, "Breyanzi" = "Liso-cel", "Yescarta" = "Axi-cel")
  ) %>%
  transmute(
    "ALT max (U/L)" = ALT.max,
    "AST max (U/L)" = AST.max,
    "CRP max (mg/L)" = CRP.max,
    "D-dimer max (mg/L)" = ddimer.max,
    "Ferritin max (ng/mL)" = ferritin.max,
    "IL6 max (pg/mL)" = IL6.max,
    "LDH max (U/L)" = LDH.max,
    CRS.grade = recode(CRS.grade, "0" = "0", "1" = "1-2", "2" = "1-2", "3" = "3+", "4" = "3+")
  ) %>%
  tbl_summary(
    by = CRS.grade,
    missing = "no",
    statistic = list(
      all_continuous() ~ "{median} ({p25}, {p75})",
      all_categorical() ~ "{n} ({p}%)"
    )
  ) %>%
  modify_spanning_header(all_stat_cols() ~ "**CRS grade**") %>%
  modify_header(all_stat_cols() ~ "**{level}**<br>N = {n} ({style_percent(p)}%)") %>%
  bold_labels() %>%
  add_p() %>%
  add_stat_label()

Characteristic	CRS grade			p-value1
	0 N = 29 (19%)	1-2 N = 111 (74%)	3+ N = 9 (6.0%)
ALT max (U/L), Median (IQR)	23 (20, 31)	57 (33, 85)	68 (44, 158)	<0.001
AST max (U/L), Median (IQR)	23 (20, 30)	36 (27, 58)	57 (42, 83)	<0.001
CRP max (mg/L), Median (IQR)	18 (6, 45)	114 (62, 168)	160 (123, 203)	<0.001
D-dimer max (mg/L), Median (IQR)	1.0 (0.6, 2.5)	2.3 (1.4, 3.7)	9.2 (8.1, 15.1)	<0.001
Ferritin max (ng/mL), Median (IQR)	283 (174, 491)	1,217 (556, 2,113)	11,312 (8,863, 14,982)	<0.001
IL6 max (pg/mL), Median (IQR)	20 (13, 54)	174 (56, 739)	7,125 (3,064, 21,316)	<0.001
LDH max (U/L), Median (IQR)	175 (151, 226)	250 (193, 351)	708 (594, 1,163)	<0.001
1 Kruskal-Wallis rank sum test

Show code

dataset %>%
  mutate(
    CAR.T.Product.Type = recode(CAR.T.Product.Type, "Breyanzi" = "Liso-cel", "Yescarta" = "Axi-cel")
  ) %>%
  transmute(
    "ALT max (U/L)" = ALT.max,
    "AST max (U/L)" = AST.max,
    "CRP max (mg/L)" = CRP.max,
    "D-dimer max (mg/L)" = ddimer.max,
    "Ferritin max (ng/mL)" = ferritin.max,
    "IL6 max (pg/mL)" = IL6.max,
    "LDH max (U/L)" = LDH.max,
    ICANS.grade = recode(ICANS.grade, "0" = "0", "1" = "1-2", "2" = "1-2", "3" = "3+", "4" = "3+")
  ) %>%
  tbl_summary(
    by = ICANS.grade,
    missing = "no",
    statistic = list(
      all_continuous() ~ "{median} ({p25}, {p75})",
      all_categorical() ~ "{n} ({p}%)"
    )
  ) %>%
  modify_spanning_header(all_stat_cols() ~ "**ICANS grade**") %>%
  modify_header(all_stat_cols() ~ "**{level}**<br>N = {n} ({style_percent(p)}%)") %>%
  bold_labels() %>%
  add_p() %>%
  add_stat_label()

Characteristic	ICANS grade			p-value1
	0 N = 83 (56%)	1-2 N = 39 (26%)	3+ N = 27 (18%)
ALT max (U/L), Median (IQR)	31 (23, 58)	62 (41, 88)	89 (52, 158)	<0.001
AST max (U/L), Median (IQR)	27 (21, 39)	44 (34, 59)	70 (31, 93)	<0.001
CRP max (mg/L), Median (IQR)	62 (19, 116)	123 (66, 191)	148 (110, 198)	<0.001
D-dimer max (mg/L), Median (IQR)	1.6 (0.8, 2.6)	2.5 (1.6, 3.7)	6.2 (3.3, 11.6)	<0.001
Ferritin max (ng/mL), Median (IQR)	530 (261, 1,327)	1,619 (868, 2,579)	4,174 (1,631, 9,710)	<0.001
IL6 max (pg/mL), Median (IQR)	54 (23, 154)	223 (111, 1,168)	1,454 (686, 5,459)	<0.001
LDH max (U/L), Median (IQR)	218 (165, 280)	231 (189, 391)	357 (248, 623)	<0.001
1 Kruskal-Wallis rank sum test

Table for cardiac/BP results

Show code

theme_gtsummary_compact()

dataset %>%
  mutate(
    CAR.T.Product.Type = recode(CAR.T.Product.Type,"Breyanzi"="Liso-cel","Yescarta"="Axi-cel")
  ) %>%
  transmute(
    "Product" = CAR.T.Product.Type,
    "Pre-LD SBP (mmHg)" = SBP.preLD,
    "Pre-LD DBP (mmHg)" = DBP.preLD,
    "Pre-LD MAP (mmHg)" = MAP.preLD,
    "Max SBP drop (mmHg)" = SBP.preLD - SBP.min,
    "Max DBP drop (mmHg)" = DBP.preLD - DBP.min,
    "Max MAP drop (mmHg)" = MAP.preLD - MAP.min,
    "Positive troponin (\u22650.03 ng/mL)" = ifelse(!is.na(trop.max) & trop.max >=0.03,1,0) 
  ) %>%
  tbl_summary(
    by = "Product",
    missing = "no",
    #sort = all_categorical() ~ "frequency",
    statistic = list(
      all_continuous() ~ "{median} ({p25}, {p75})",
      all_categorical() ~ "{n} ({p}%)"
    )
  ) %>%
  bold_labels %>%
  add_p() %>%
  add_stat_label() %>%
  sort_p()

Characteristic	Axi-cel, N = 99	Liso-cel, N = 50	p-value1
Max DBP drop (mmHg), Median (IQR)	27 (20, 34)	20 (14, 25)	<0.001
Max MAP drop (mmHg), Median (IQR)	29 (23, 36)	23 (17, 28)	<0.001
Max SBP drop (mmHg), Median (IQR)	35 (28, 45)	27 (19, 37)	0.001
Pre-LD SBP (mmHg), Median (IQR)	116 (108, 127)	116 (111, 123)	0.6
Positive troponin (≥0.03 ng/mL), n (%)	4 (4.0%)	3 (6.0%)	0.7
Pre-LD MAP (mmHg), Median (IQR)	87 (81, 95)	88 (83, 94)	0.8
Pre-LD DBP (mmHg), Median (IQR)	74 (67, 80)	74 (69, 77)	0.8
1 Wilcoxon rank sum test; Fisher’s exact test

STACKED BAR GRAPH

Stacked bar graph of toxicity for liso-cel and axi-cel

Show code

combined_data <- rbind(
  dataset %>%
    count(CAR.T.Product.Type, CRS.grade) %>%
    group_by(CAR.T.Product.Type) %>%
    mutate(
      total = sum(n[CRS.grade != 0]),
      pct = prop.table(n) * 100,
      total_pct = sum(pct[CRS.grade != 0]),
      Grade = factor(CRS.grade, levels = c(0, 4, 3, 2, 1)),
      CAR.T.Product.Type = recode(CAR.T.Product.Type, "Breyanzi" = "Liso-cel", "Yescarta" = "Axi-cel"),
      Category = "CRS"
    ),
  dataset %>%
    count(CAR.T.Product.Type, ICANS.grade) %>%
    group_by(CAR.T.Product.Type) %>%
    mutate(
      total = sum(n[ICANS.grade != 0]),
      pct = prop.table(n) * 100,
      total_pct = sum(pct[ICANS.grade != 0]),
      Grade = factor(ICANS.grade, levels = c(0, 4, 3, 2, 1)),
      CAR.T.Product.Type = recode(CAR.T.Product.Type, "Breyanzi" = "Liso-cel", "Yescarta" = "Axi-cel"),
      Category = "ICANS"
    )
)

ggplot(combined_data) +
  aes(x = CAR.T.Product.Type, y = pct, fill = Grade) +
  geom_bar(stat = "identity", position = "stack", data = . %>% filter(Grade != 0)) +
    geom_text(aes(label = paste0(sprintf("%1.1f", pct), "% (",n,")")), 
            data = . %>% filter(Grade != 0),
            position = position_stack(vjust = 0.5)) + 
  geom_text( aes(label = paste0(sprintf("%1.1f", total_pct), "% (",total,")"), x = CAR.T.Product.Type, y = total_pct + 7, vjust = 0), 
             data = . %>% filter(Grade != 0),
             color = "#104a8e") +
  #geom_text(aes(label = paste0(sprintf("%1.1f", pct), "% (", n, ")")), position = position_dodge(width = 0.9), vjust = 0.5) +
  #geom_text(aes(label = paste0(sprintf("%1.1f", total_pct), "% (", total, ")"), y = total_pct + 7), color = "#104a8e") +
  facet_wrap(~Category, scales = "free_y") +
  ylab("Proportion of patients (%)") +
  xlab(NULL) +
  #ggtitle("Comparison of CRS and ICANS grades") +
  theme_classic()+
  scale_fill_brewer(palette = "Pastel1") +
  scale_y_continuous(breaks = seq(0, 100, by = 20), limits = c(0, 100)) +
  theme(
    axis.title = element_text(size = 16),
    axis.text = element_text(size = 16),
    legend.title = element_text(size = 16),
    legend.text = element_text(size = 14),
    axis.text.x = element_text(angle = 45, hjust = 1)
  )

COMPETING RISK ANALYSIS

Cumulative incidence of death

Show code

theme_gtsummary_compact()

data1 <- filter(dataset, Infused == 1) %>%
  select(Days.to.DLC, Alive.RM.NRM, CAR.T.Product.Type, Age, Largest.lesion, ALC.preLD, HCT.CI, LDH.preLD) %>%
  mutate(
   Alive.RM.NRM = as.factor(recode(Alive.RM.NRM,"0"="Alive","1"="Death from cancer","2"="Death from other causes")),
   CAR.T.Product.Type = factor(recode(CAR.T.Product.Type, "Breyanzi" = "Liso-cel", "Yescarta" = "Axi-cel"), levels = c("Liso-cel","Axi-cel") ),
   Months.to.DLC = Days.to.DLC/30
  )

cum <- tidycmprsk::cuminc(Surv( Months.to.DLC , Alive.RM.NRM) ~ CAR.T.Product.Type, data1, rho=0, conf.level=0.95)

cum %>%
  ggcuminc(
   #risk.table=TRUE,
   theme = theme_classic(),
   outcome = c("Death from cancer","Death from other causes")
  ) +
  labs(
    x = "Time (months)",
    y = "Probability of death"
  ) +
  ylim(0,1) +
  scale_x_continuous(breaks = seq(0, 36, by = 6), limits=c(0, 36)) +
  #add_confidence_interval() + 
  add_risktable(
    size = 5,
    risktable_stats = c("n.risk"),
    risktable_height = 0.2,
    stats_label = c("Number at risk"),
    theme = theme_risktable_default(axis.text.y.size = 14, plot.title.size = 14)
  )+ 
  add_pvalue() + 
  theme(
    legend.position =  c(0.22, 0.8),
    text = element_text(size = 16)  # Set text size to 16
  )

Show code

cum %>%
  tbl_cuminc(
    times = c(12),
    outcomes = c("Death from cancer","Death from other causes"),
    label_header = c("**{time}-month cuminc**")
  ) %>%
  add_p()

Characteristic	12-month cuminc	p-value1
Death from cancer
CAR.T.Product.Type		0.5
Liso-cel	27% (14%, 42%)
Axi-cel	39% (28%, 50%)
Death from other causes
CAR.T.Product.Type		>0.9
Liso-cel	7.3% (1.8%, 18%)
Axi-cel	5.2% (1.7%, 12%)
1 Gray’s Test

Show code

uv_tab<-crr(Surv(Days.to.DLC, Alive.RM.NRM) ~ CAR.T.Product.Type, data = data1) %>%
  tbl_regression(exp = TRUE)

mv_tab<-crr(Surv(Days.to.DLC, Alive.RM.NRM) ~ CAR.T.Product.Type + Age + Largest.lesion + ALC.preLD + LDH.preLD, data = data1) %>%
  tbl_regression(exp = TRUE)

1 cases omitted due to missing values

Show code

tbl_merge(
  tbls = list(uv_tab, mv_tab),                          # combine
  tab_spanner = c("**Univariate**", "**Multivariable**")) # set header names

Characteristic	Univariate			Multivariable
	HR1	95% CI1	p-value	HR1	95% CI1	p-value
CAR.T.Product.Type
Liso-cel	—	—		—	—
Axi-cel	1.22	0.70, 2.13	0.5	1.10	0.61, 1.99	0.8
Age				1.00	0.98, 1.02	0.8
Largest.lesion				1.10	1.03, 1.18	0.007
ALC.preLD				1.31	1.20, 1.43	<0.001
LDH.preLD				1.00	1.00, 1.00	<0.001
1 HR = Hazard Ratio, CI = Confidence Interval

KAPLAN-MEIER ANALYSIS

Calculate length of follow-up

Show code

library(prodlim)
library(haven)

data1 = dataset %>%
  mutate(
    Reverse_death = ifelse(Death == 1, 0,1)
  )

#quantile(prodlim(Hist(time = Days.to.DLC/30, Death )~1, data = data1, reverse = TRUE ) )

reverse_km_OS <- survfit(Surv(Days.to.DLC/30, Reverse_death) ~ 1, data1)
reverse_km_OS

Call: survfit(formula = Surv(Days.to.DLC/30, Reverse_death) ~ 1, data = data1)

       n events median 0.95LCL 0.95UCL
[1,] 149     83   27.1    22.6    38.2

Show code

reverse_km_OS_breyanzi <- survfit(Surv(Days.to.DLC/30, Reverse_death) ~ 1, filter(data1, CAR.T.Product.Type == "Breyanzi") )
reverse_km_OS_breyanzi

Call: survfit(formula = Surv(Days.to.DLC/30, Reverse_death) ~ 1, data = filter(data1, 
    CAR.T.Product.Type == "Breyanzi"))

      n events median 0.95LCL 0.95UCL
[1,] 50     30   18.5    12.9    25.9

Show code

reverse_km_OS_yescarta <- survfit(Surv(Days.to.DLC/30, Reverse_death) ~ 1, filter(data1, CAR.T.Product.Type == "Yescarta") )
reverse_km_OS_yescarta

Call: survfit(formula = Surv(Days.to.DLC/30, Reverse_death) ~ 1, data = filter(data1, 
    CAR.T.Product.Type == "Yescarta"))

      n events median 0.95LCL 0.95UCL
[1,] 99     53     43    32.5    51.5

KM for OS, whole cohort

Show code

data1 = filter(dataset, Infused == 1)

km_OS <- survfit(Surv(Days.to.DLC/30, Death) ~ 1, data1)
km_OS

Call: survfit(formula = Surv(Days.to.DLC/30, Death) ~ 1, data = data1)

       n events median 0.95LCL 0.95UCL
[1,] 149     66   16.3    12.1    53.5

Show code

#autoplot(km_Regain)
OS_plot <- ggsurvplot(km_OS,
           pval=FALSE,
           conf.int = TRUE,
           risk.table = TRUE, # Add risk table
           fontsize = 6,
           risk.table.col = "strata", # Change risk table color by groups
           tables.height = 0.25,
           tables.theme = theme_cleantable(),
           xlab="Time (months)", ylab = "Overall survival",
           xlim = c(0,36), break.x.by = c(180/30),
           ylim = c(0,1), break.y.by = c(0.25),
           legend = "none",
           surv.scale="percent",
           font.main = c(16, "plain", "black"),
           font.x = c(16, "plain", "black"),
           font.y = c(16, "plain", "black"),
           font.caption = c(16, "plain", "black"),
           font.tickslab = c(16, "plain", "black")
) 

OS_plot

KM for OS, whole cohort

Show code

data1 = filter(dataset, Infused == 1)

km_PFS <- survfit(Surv(Days.to.relapse.death.or.DLC/30, Relapse.or.death) ~ 1, data1)
km_PFS

Call: survfit(formula = Surv(Days.to.relapse.death.or.DLC/30, Relapse.or.death) ~ 
    1, data = data1)

       n events median 0.95LCL 0.95UCL
[1,] 149     84   7.43    5.33    19.6

Show code

#autoplot(km_Regain)
PFS_plot <- ggsurvplot(km_PFS,
           pval=FALSE,
           conf.int = TRUE,
           risk.table = TRUE, # Add risk table
           fontsize = 6,
           risk.table.col = "strata", # Change risk table color by groups
           tables.height = 0.25,
           tables.theme = theme_cleantable(),
           xlab="Time (months)", ylab = "Progression-free survival",
           xlim = c(0,36), break.x.by = c(180/30),
           ylim = c(0,1), break.y.by = c(0.25),
           legend = "none",
           surv.scale="percent",
           font.main = c(16, "plain", "black"),
           font.x = c(16, "plain", "black"),
           font.y = c(16, "plain", "black"),
           font.caption = c(16, "plain", "black"),
           font.tickslab = c(16, "plain", "black")
) 

PFS_plot

Create KM for OS

Show code

data1 = filter(dataset, Infused == 1) %>%
  mutate(
    Reverse_death = ifelse(Death == 1, 0,1)
  )

km_OS <- survfit(Surv(Days.to.DLC/30, Death) ~ CAR.T.Product.Type, data1)
km_OS

Call: survfit(formula = Surv(Days.to.DLC/30, Death) ~ CAR.T.Product.Type, 
    data = data1)

                             n events median 0.95LCL 0.95UCL
CAR.T.Product.Type=Breyanzi 50     20   15.6    13.2      NA
CAR.T.Product.Type=Yescarta 99     46   16.0     9.4      NA

Show code

med_OS <- surv_median(km_OS)

p.logrank <- survdiff(Surv(Days.to.DLC/30, Death) ~ CAR.T.Product.Type, data1)$pvalue


oneYr <- summary(km_OS,times=c(12))
oneYr

Call: survfit(formula = Surv(Days.to.DLC/30, Death) ~ CAR.T.Product.Type, 
    data = data1)

                CAR.T.Product.Type=Breyanzi 
        time       n.risk      n.event     survival      std.err 
      12.000       21.000       14.000        0.655        0.077 
lower 95% CI upper 95% CI 
       0.520        0.825 

                CAR.T.Product.Type=Yescarta 
        time       n.risk      n.event     survival      std.err 
     12.0000      41.0000      35.0000       0.5543       0.0565 
lower 95% CI upper 95% CI 
      0.4540       0.6769 

Show code

oneYr$surv[1]- oneYr$surv[2] # Difference in survival at 1 yr for Breyanzi vs Yescarta

[1] 0.1008576

Show code

print("Difference in survival: ")

[1] "Difference in survival: "

Show code

diffSE <- sqrt(oneYr$std.err[2]^2 + oneYr$std.err[1]^2) # calculate the standard error

#Calculate the 95% CI for the difference
print("95% CI: ")

[1] "95% CI: "

Show code

oneYr$surv[1] - oneYr$surv[2] - 1.96 * diffSE

[1] -0.08635184

Show code

oneYr$surv[1] - oneYr$surv[2] + 1.96 * diffSE

[1] 0.288067

Show code

zStat <- (oneYr$surv[1] - oneYr$surv[2])/diffSE # calculate the c-test statistic

print("p-value: ")

[1] "p-value: "

Show code

p = 2 * pnorm(abs(zStat), lower.tail = FALSE) # calculate a two-sided p-value


#autoplot(km_Regain)
OS_plot <- ggsurvplot(km_OS,
           pval=FALSE,
           conf.int = TRUE,
           risk.table = TRUE, # Add risk table
           fontsize = 6,
           risk.table.col = "strata", # Change risk table color by groups
           tables.height = 0.25,
           tables.theme = theme_cleantable(),
           xlab="Time (months)", ylab = "Overall survival",
           xlim = c(0,36), break.x.by = c(180/30),
           ylim = c(0,1), break.y.by = c(0.25),
           legend.labs = c("Liso-cel","Axi-cel"),
           legend = "none",
           surv.scale="percent",
           font.main = c(16, "plain", "black"),
           font.x = c(16, "plain", "black"),
           font.y = c(16, "plain", "black"),
           font.caption = c(16, "plain", "black"),
           font.tickslab = c(16, "plain", "black")
) 

OS_plot$plot <- OS_plot$plot + 
  geom_vline(xintercept = 12, linetype = "dashed", color = "steelblue") +
  geom_hline(yintercept = 0.5, linetype = "dashed", color = "black") +
  annotate("text", x = 12.5, y = 0.83, label = paste("Liso-cel vs axi-cel: 1-year OS\n", signif(oneYr$surv[1]*100, 2), "% vs ", signif(oneYr$surv[2]*100, 2), "%, p =", signif(p,2)), color = "steelblue", hjust = 0, vjust = -0.5, size = 3) +
  annotate("text", x = 13, y = 0, label = paste("Liso-cel vs axi-cel,\nmedian OS:\n", signif(med_OS$median[1], 2), "vs", signif(med_OS$median[2], 2), "months, p =", signif(p.logrank,2)), color = "black", hjust = 0, vjust = -0.5, size = 3, fontface = "bold")


OS_plot

Create KM for PFS, stratified by product

Show code

data1 <- filter(dataset, Infused == 1) #%>%
  #filter(Best.response == "CR" | Best.response == "PR") ## To view DOR instead of PFS

km_RFS <- survfit(Surv(Days.to.relapse.death.or.DLC/30, Relapse.or.death) ~ CAR.T.Product.Type, data = data1 )
km_RFS

Call: survfit(formula = Surv(Days.to.relapse.death.or.DLC/30, Relapse.or.death) ~ 
    CAR.T.Product.Type, data = data1)

                             n events median 0.95LCL 0.95UCL
CAR.T.Product.Type=Breyanzi 50     30   7.17    4.07    22.6
CAR.T.Product.Type=Yescarta 99     54   9.50    4.57    41.1

Show code

med_RFS <- surv_median(km_RFS)

p.logrank <- survdiff(Surv(Days.to.relapse.death.or.DLC/30, Relapse.or.death) ~ CAR.T.Product.Type, data1)$pvalue


oneYr <- summary(km_RFS,times=c(12))
oneYr

Call: survfit(formula = Surv(Days.to.relapse.death.or.DLC/30, Relapse.or.death) ~ 
    CAR.T.Product.Type, data = data1)

                CAR.T.Product.Type=Breyanzi 
        time       n.risk      n.event     survival      std.err 
     12.0000      12.0000      27.0000       0.3839       0.0776 
lower 95% CI upper 95% CI 
      0.2584       0.5705 

                CAR.T.Product.Type=Yescarta 
        time       n.risk      n.event     survival      std.err 
     12.0000      34.0000      46.0000       0.4591       0.0556 
lower 95% CI upper 95% CI 
      0.3621       0.5821 

Show code

oneYr$surv[1]- oneYr$surv[2] # Difference in survival at 1 yr for Breyanzi vs Yescarta

[1] -0.07517689

Show code

print("Difference in survival: ")

[1] "Difference in survival: "

Show code

diffSE <- sqrt(oneYr$std.err[2]^2 + oneYr$std.err[1]^2) # calculate the standard error

#Calculate the 95% CI for the difference
print("95% CI: ")

[1] "95% CI: "

Show code

oneYr$surv[1] - oneYr$surv[2] - 1.96 * diffSE

[1] -0.2622901

Show code

oneYr$surv[1] - oneYr$surv[2] + 1.96 * diffSE

[1] 0.1119363

Show code

zStat <- (oneYr$surv[1] - oneYr$surv[2])/diffSE # calculate the c-test statistic

print("p-value: ")

[1] "p-value: "

Show code

p = 2 * pnorm(abs(zStat), lower.tail = FALSE) # calculate a two-sided p-value


#autoplot(km_Regain)
RFS_plot <- ggsurvplot(km_RFS,
           pval=FALSE,
           conf.int = TRUE,
           risk.table = TRUE, # Add risk table
           fontsize = 6,
           risk.table.col = "strata", # Change risk table color by groups
           tables.height = 0.25,
           tables.theme = theme_cleantable(),
           xlab="Time (months)", ylab = "Progression-free survival",
           xlim = c(0,36), break.x.by = c(180/30),
           ylim = c(0,1), break.y.by = c(0.25),
           legend.labs = c("Liso-cel","Axi-cel"),
           legend = "none",
           surv.scale="percent",
           font.main = c(16, "plain", "black"),
           font.x = c(16, "plain", "black"),
           font.y = c(16, "plain", "black"),
           font.caption = c(16, "plain", "black"),
           font.tickslab = c(16, "plain", "black")
) 

RFS_plot$plot <- RFS_plot$plot + 
  geom_vline(xintercept = 12, linetype = "dashed", color = "steelblue") +
  geom_hline(yintercept = 0.5, linetype = "dashed", color = "black") +
  annotate("text", x = 12.5, y = 0.83, label = paste("Liso-cel vs axi-cel: 1-year PFS\n", signif(oneYr$surv[1]*100, 2), "% vs ", signif(oneYr$surv[2]*100, 2), "%, p =", signif(p,2)), color = "steelblue", hjust = 0, vjust = -0.5, size = 3) +
  annotate("text", x = 13, y = .5, label = paste("Liso-cel vs axi-cel,\nmedian PFS:\n", signif(med_RFS$median[1], 2), "vs", signif(med_RFS$median[2], 2), "months, p =", signif(p.logrank,2)), color = "black", hjust = 0, vjust = -0.5, size = 3, fontface = "bold")


RFS_plot

Create KM for DOR, stratified by product

Show code

data1 <- filter(dataset, Infused == 1) %>%
  filter(Best.response == "CR" | Best.response == "PR") ## To view DOR instead of PFS

km_DOR <- survfit(Surv( (Days.to.relapse.death.or.DLC-Days.to.response)/30, Relapse.or.death) ~ CAR.T.Product.Type, data = data1 )
km_DOR

Call: survfit(formula = Surv((Days.to.relapse.death.or.DLC - Days.to.response)/30, 
    Relapse.or.death) ~ CAR.T.Product.Type, data = data1)

                             n events median 0.95LCL 0.95UCL
CAR.T.Product.Type=Breyanzi 40     20   11.0    5.83      NA
CAR.T.Product.Type=Yescarta 76     31   40.2    9.93      NA

Show code

med_DOR <- surv_median(km_DOR)

p.logrank <- survdiff(Surv( (Days.to.relapse.death.or.DLC-Days.to.response)/30, Relapse.or.death) ~ CAR.T.Product.Type, data1)$pvalue


oneYr <- summary(km_DOR,times=c(12))
oneYr

Call: survfit(formula = Surv((Days.to.relapse.death.or.DLC - Days.to.response)/30, 
    Relapse.or.death) ~ CAR.T.Product.Type, data = data1)

                CAR.T.Product.Type=Breyanzi 
        time       n.risk      n.event     survival      std.err 
     12.0000      12.0000      17.0000       0.4830       0.0906 
lower 95% CI upper 95% CI 
      0.3344       0.6977 

                CAR.T.Product.Type=Yescarta 
        time       n.risk      n.event     survival      std.err 
     12.0000      34.0000      23.0000       0.6054       0.0643 
lower 95% CI upper 95% CI 
      0.4916       0.7455 

Show code

oneYr$surv[1]- oneYr$surv[2] # Difference in survival at 1 yr for Breyanzi vs Yescarta

[1] -0.1223713

Show code

print("Difference in survival: ")

[1] "Difference in survival: "

Show code

diffSE <- sqrt(oneYr$std.err[2]^2 + oneYr$std.err[1]^2) # calculate the standard error

#Calculate the 95% CI for the difference
print("95% CI: ")

[1] "95% CI: "

Show code

oneYr$surv[1] - oneYr$surv[2] - 1.96 * diffSE

[1] -0.3401687

Show code

oneYr$surv[1] - oneYr$surv[2] + 1.96 * diffSE

[1] 0.09542613

Show code

zStat <- (oneYr$surv[1] - oneYr$surv[2])/diffSE # calculate the c-test statistic

print("p-value: ")

[1] "p-value: "

Show code

p = 2 * pnorm(abs(zStat), lower.tail = FALSE) # calculate a two-sided p-value


#autoplot(km_Regain)
DOR_plot <- ggsurvplot(km_DOR,
           pval=FALSE,
           conf.int = TRUE,
           risk.table = TRUE, # Add risk table
           fontsize = 6,
           risk.table.col = "strata", # Change risk table color by groups
           tables.height = 0.25,
           tables.theme = theme_cleantable(),
           xlab="Time (months)", ylab = "Progression-free survival",
           xlim = c(0,36), break.x.by = c(180/30),
           ylim = c(0,1), break.y.by = c(0.25),
           legend.labs = c("Liso-cel","Axi-cel"),
           legend = "none",
           surv.scale="percent",
           font.main = c(16, "plain", "black"),
           font.x = c(16, "plain", "black"),
           font.y = c(16, "plain", "black"),
           font.caption = c(16, "plain", "black"),
           font.tickslab = c(16, "plain", "black")
) 

DOR_plot$plot <- DOR_plot$plot + 
  geom_vline(xintercept = 12, linetype = "dashed", color = "steelblue") +
  geom_hline(yintercept = 0.5, linetype = "dashed", color = "black") +
  annotate("text", x = 12.5, y = 0.83, label = paste("Liso-cel vs axi-cel: 1-year DOR\n", signif(oneYr$surv[1]*100, 2), "% vs ", signif(oneYr$surv[2]*100, 2), "%, p =", signif(p,2)), color = "steelblue", hjust = 0, vjust = -0.5, size = 3) +
  annotate("text", x = 13, y = .5, label = paste("Liso-cel vs axi-cel,\nmedian DOR:\n", signif(med_DOR$median[1], 2), "vs", signif(med_DOR$median[2], 2), "months, p =", signif(p.logrank,2)), color = "black", hjust = 0, vjust = -0.5, size = 3, fontface = "bold")


DOR_plot

Combined KM with OS and PFS

Show code

splots <- list()
splots[[1]] = OS_plot
splots[[2]] = RFS_plot
arrange_ggsurvplots(splots, print = TRUE,
  ncol = 2, nrow = 1, risk.table.height = 0.25)

Create KM for OS stratified by best response

Show code

km_OS <- survfit(Surv(Days.to.DLC/30, Death) ~ Best.response, data = filter(dataset, Infused == 1 & Best.response != "SD"))
km_OS

Call: survfit(formula = Surv(Days.to.DLC/30, Death) ~ Best.response, 
    data = filter(dataset, Infused == 1 & Best.response != "SD"))

                  n events median 0.95LCL 0.95UCL
Best.response=CR 77     21     NA   53.50      NA
Best.response=PD 32     25    5.5    3.67     9.7
Best.response=PR 39     19   13.5    6.77      NA

Show code

#autoplot(km_Regain)

OS_plot <- ggsurvplot(km_OS,
           pval=TRUE,
           conf.int = TRUE,
           risk.table = TRUE, # Add risk table
           fontsize = 6,
           risk.table.col = "strata", # Change risk table color by groups
           tables.height = 0.3,
           tables.theme = theme_cleantable(),
           #surv.median.line = "hv", # Specify median survival
           #ggtheme = theme_bw(), # Change ggplot2 theme
           #palette = c("#E7B800", "#2E9FDF"),
           xlab="Time (months)", ylab = "Overall survival",
           xlim = c(0,72), break.x.by = c(12),
           ylim = c(0,1), break.y.by = c(0.25),
           legend.labs = c("CR","PD", "PR"),
           legend = "none",
           surv.scale="percent",
           font.main = c(16, "plain", "black"),
           font.x = c(16, "plain", "black"),
           font.y = c(16, "plain", "black"),
           font.caption = c(16, "plain", "black"),
           font.tickslab = c(16, "plain", "black")
)
OS_plot

Create KM for OS stratified by Bulky disease (>=5cm) status

Show code

km_OS <- survfit(Surv(Days.to.DLC/30, Death) ~ Bulky, data = filter(dataset, Infused == 1 | Best.response != "") )
km_OS

Call: survfit(formula = Surv(Days.to.DLC/30, Death) ~ Bulky, data = filter(dataset, 
    Infused == 1 | Best.response != ""))

          n events median 0.95LCL 0.95UCL
Bulky=0 105     39  43.23    16.0      NA
Bulky=1  44     27   6.37     4.7    13.5

Show code

#autoplot(km_Regain)
OS_plot <- ggsurvplot(km_OS,
           pval=TRUE,
           conf.int = TRUE,
           risk.table = TRUE, # Add risk table
           fontsize = 6,
           risk.table.col = "strata", # Change risk table color by groups
           tables.height = 0.25,
           tables.theme = theme_cleantable(),
           xlab="Time (months)", ylab = "Overall survival",
           xlim = c(0,72), break.x.by = c(12),
           ylim = c(0,1), break.y.by = c(0.25),
           legend.labs = c("Non-bulky","Bulky"),
           legend = "none",
           surv.scale="percent",
           font.main = c(16, "plain", "black"),
           font.x = c(16, "plain", "black"),
           font.y = c(16, "plain", "black"),
           font.caption = c(16, "plain", "black"),
           font.tickslab = c(16, "plain", "black")
)
OS_plot

Create KM for OS stratified by Breyanzi out-of-spec (EAP) status

Show code

km_OS <- survfit(Surv(Days.to.DLC, Death) ~ EAP, data = filter(dataset, CAR.T.Product.Type == "Breyanzi") )
km_OS

Call: survfit(formula = Surv(Days.to.DLC, Death) ~ EAP, data = filter(dataset, 
    CAR.T.Product.Type == "Breyanzi"))

       n events median 0.95LCL 0.95UCL
EAP=0 39     16    409     357      NA
EAP=1 11      4     NA     252      NA

Show code

OS_plot <- ggsurvplot(km_OS,
           pval=TRUE,
           conf.int = TRUE,
           risk.table = TRUE, # Add risk table
           fontsize = 6,
           risk.table.col = "strata", # Change risk table color by groups
           tables.height = 0.25,
           tables.theme = theme_cleantable(),
           xlab="Time (days)", ylab = "Overall survival",
           xlim = c(0,1080), break.x.by = c(180),
           ylim = c(0,1), break.y.by = c(0.25),
           legend.labs = c("In-spec", "Out-of-spec"),
           legend = "none",
           surv.scale="percent",
           font.main = c(16, "plain", "black"),
           font.x = c(16, "plain", "black"),
           font.y = c(16, "plain", "black"),
           font.caption = c(16, "plain", "black"),
           font.tickslab = c(16, "plain", "black")
)
OS_plot

Create KM for OS stratified by LDH status (<210 vs >=210 U/L)

Show code

data1 = filter(dataset, Infused == 1)

km_OS <- survfit(Surv(Days.to.DLC/30, Death) ~ High.LDH, data1)
km_OS

Call: survfit(formula = Surv(Days.to.DLC/30, Death) ~ High.LDH, data = data1)

   1 observation deleted due to missingness 
            n events median 0.95LCL 0.95UCL
High.LDH=0 84     29  43.23   24.30      NA
High.LDH=1 64     36   6.77    4.57    11.9

Show code

#autoplot(km_Regain)


OS_plot <- ggsurvplot(km_OS,
           pval=TRUE,
           conf.int = TRUE,
           risk.table = TRUE, # Add risk table
           fontsize = 6,
           risk.table.col = "strata", # Change risk table color by groups
           tables.height = 0.25,
           tables.theme = theme_cleantable(),
           #surv.median.line = "hv", # Specify median survival
           #ggtheme = theme_bw(), # Change ggplot2 theme
           #palette = c("#E7B800", "#2E9FDF"),
           xlab="Time (months)", ylab = "Overall survival",
           xlim = c(0,72), break.x.by = c(12),
           ylim = c(0,1), break.y.by = c(0.25),
           legend.labs = c("LDH <210 U/L", "LDH \u2265210 U/L"),
           legend = "none",
           surv.scale="percent",
           font.main = c(16, "plain", "black"),
           font.x = c(16, "plain", "black"),
           font.y = c(16, "plain", "black"),
           font.caption = c(16, "plain", "black"),
           font.tickslab = c(16, "plain", "black")
)
OS_plot

Create KM for OS stratified by HCT-CI status (>=3 vs <3)

Show code

data1 = filter(dataset, Infused == 1) %>%
  mutate(
    High.HCT.CI = ifelse(HCT.CI >=3,1,0) )

km_OS <- survfit(Surv(Days.to.DLC/30, Death) ~ High.HCT.CI, data1 )
km_OS

Call: survfit(formula = Surv(Days.to.DLC/30, Death) ~ High.HCT.CI, 
    data = data1)

                n events median 0.95LCL 0.95UCL
High.HCT.CI=0 100     42   16.3   13.07      NA
High.HCT.CI=1  49     24   16.0    8.97      NA

Show code

OS_plot <- ggsurvplot(km_OS,
           pval=TRUE,
           conf.int = TRUE,
           risk.table = TRUE, # Add risk table
           fontsize = 6,
           risk.table.col = "strata", # Change risk table color by groups
           tables.height = 0.25,
           tables.theme = theme_cleantable(),
           #surv.median.line = "hv", # Specify median survival
           #ggtheme = theme_bw(), # Change ggplot2 theme
           #palette = c("#E7B800", "#2E9FDF"),
           xlab="Time (months)", ylab = "Overall survival",
           xlim = c(0,72), break.x.by = c(12),
           ylim = c(0,1), break.y.by = c(0.25),
           legend.labs = c("HCT <3", "HCT \u22653"),
           legend = "none",
           surv.scale="percent",
           font.main = c(16, "plain", "black"),
           font.x = c(16, "plain", "black"),
           font.y = c(16, "plain", "black"),
           font.caption = c(16, "plain", "black"),
           font.tickslab = c(16, "plain", "black")
)
OS_plot

Create KM for OS stratified by bridging response category

Show code

data1 = dataset %>%
  mutate(
    Bridging.response.category = factor(recode(Bridging.response.category, "0" = "No bridging", "1" = "CR/PR", "2" = "SD/PD"), levels = c("No bridging","CR/PR","SD/PD"))
    )

km_OS <- survfit(Surv(Days.to.DLC/30, Death) ~ Bridging.response.category, data1 )
km_OS

Call: survfit(formula = Surv(Days.to.DLC/30, Death) ~ Bridging.response.category, 
    data = data1)

   4 observations deleted due to missingness 
                                        n events median 0.95LCL
Bridging.response.category=No bridging 88     32  43.97   16.00
Bridging.response.category=CR/PR       21      9  13.57    7.07
Bridging.response.category=SD/PD       36     23   6.53    4.70
                                       0.95UCL
Bridging.response.category=No bridging      NA
Bridging.response.category=CR/PR            NA
Bridging.response.category=SD/PD          13.6

Show code

OS_plot <- ggsurvplot(km_OS,
           pval=TRUE,
           conf.int = TRUE,
           risk.table = TRUE, # Add risk table
           fontsize = 6,
           risk.table.col = "strata", # Change risk table color by groups
           tables.height = 0.25,
           tables.theme = theme_cleantable(),
           #surv.median.line = "hv", # Specify median survival
           #ggtheme = theme_bw(), # Change ggplot2 theme
           #palette = c("#E7B800", "#2E9FDF"),
           xlab="Time (months)", ylab = "Overall survival",
           xlim = c(0,72), break.x.by = c(12),
           ylim = c(0,1), break.y.by = c(0.25),
           legend.labs = c("No bridging","CR/PR","SD/PD"),
           legend = "none",
           surv.scale="percent",
           font.main = c(16, "plain", "black"),
           font.x = c(16, "plain", "black"),
           font.y = c(16, "plain", "black"),
           font.caption = c(16, "plain", "black"),
           font.tickslab = c(16, "plain", "black")
)
OS_plot

Create KM for PFS stratified by best response

Show code

km_PFS <- survfit(Surv(Days.to.relapse.death.or.DLC/30, Relapse.or.death) ~ Best.response, data = filter(dataset, Infused == 1 & Best.response != "SD"& Best.response != "PD") )
km_PFS

Call: survfit(formula = Surv(Days.to.relapse.death.or.DLC/30, Relapse.or.death) ~ 
    Best.response, data = filter(dataset, Infused == 1 & Best.response != 
    "SD" & Best.response != "PD"))

                  n events median 0.95LCL 0.95UCL
Best.response=CR 77     27     NA   22.57      NA
Best.response=PR 39     24   4.07    3.47    19.6

Show code

#autoplot(km_Regain)

PFS_plot <- ggsurvplot(km_PFS,
           pval=TRUE,
           conf.int = TRUE,
           risk.table = TRUE, # Add risk table
           fontsize = 6,
           risk.table.col = "strata", # Change risk table color by groups
           tables.height = 0.25,
           tables.theme = theme_cleantable(),
           #surv.median.line = "hv", # Specify median survival
           #ggtheme = theme_bw(), # Change ggplot2 theme
           #palette = c("#E7B800", "#2E9FDF"),
           xlab="Time (months)", ylab = "Progression-free survival",
           xlim = c(0,72), break.x.by = c(12),
           ylim = c(0,1), break.y.by = c(0.25),
           legend.labs = c("CR", "PR"),
           legend = "none",
           surv.scale="percent",
           font.main = c(16, "plain", "black"),
           font.x = c(16, "plain", "black"),
           font.y = c(16, "plain", "black"),
           font.caption = c(16, "plain", "black"),
           font.tickslab = c(16, "plain", "black")
)
PFS_plot

Create KM for PFS stratified by Bulky disease (>=5cm) status

Show code

km_PFS <- survfit(Surv(Days.to.relapse.death.or.DLC/30, Relapse.or.death) ~ Bulky, data = filter(dataset, Infused == 1 | Best.response != "") )
km_PFS

Call: survfit(formula = Surv(Days.to.relapse.death.or.DLC/30, Relapse.or.death) ~ 
    Bulky, data = filter(dataset, Infused == 1 | Best.response != 
    ""))

          n events median 0.95LCL 0.95UCL
Bulky=0 105     55  11.90    6.77      44
Bulky=1  44     29   3.63    1.47      22

Show code

#autoplot(km_Regain)
PFS_plot <- ggsurvplot(km_PFS,
           pval=TRUE,
           conf.int = TRUE,
           risk.table = TRUE, # Add risk table
           fontsize = 6,
           risk.table.col = "strata", # Change risk table color by groups
           tables.height = 0.25,
           tables.theme = theme_cleantable(),
           xlab="Time (months)", ylab = "Progression-free survival",
           xlim = c(0,72), break.x.by = c(12),
           ylim = c(0,1), break.y.by = c(0.25),
           legend.labs = c("Non-bulky","Bulky"),
           legend = "none",
           surv.scale="percent",
           font.main = c(16, "plain", "black"),
           font.x = c(16, "plain", "black"),
           font.y = c(16, "plain", "black"),
           font.caption = c(16, "plain", "black"),
           font.tickslab = c(16, "plain", "black")
)
PFS_plot

Create KM for PFS stratified by Breyanzi out-of-spec (EAP) status

Show code

km_PFS <- survfit(Surv(Days.to.DLC/30, Relapse.or.death) ~ EAP, data = filter(dataset, CAR.T.Product.Type == "Breyanzi") )
km_PFS

Call: survfit(formula = Surv(Days.to.DLC/30, Relapse.or.death) ~ EAP, 
    data = filter(dataset, CAR.T.Product.Type == "Breyanzi"))

       n events median 0.95LCL 0.95UCL
EAP=0 39     22   13.2    8.67    25.5
EAP=1 11      8   24.5    8.40      NA

Show code

PFS_plot <- ggsurvplot(km_PFS,
           pval=TRUE,
           conf.int = TRUE,
           risk.table = TRUE, # Add risk table
           fontsize = 6,
           risk.table.col = "strata", # Change risk table color by groups
           tables.height = 0.25,
           tables.theme = theme_cleantable(),
           xlab="Time (months)", ylab = "Progression-free survival",
           xlim = c(0,72), break.x.by = c(12),
           ylim = c(0,1), break.y.by = c(0.25),
           legend.labs = c("In-spec", "Out-of-spec"),
           legend = "none",
           surv.scale="percent",
           font.main = c(16, "plain", "black"),
           font.x = c(16, "plain", "black"),
           font.y = c(16, "plain", "black"),
           font.caption = c(16, "plain", "black"),
           font.tickslab = c(16, "plain", "black")
)
PFS_plot

Create KM for PFS stratified by LDH status (<210 vs >=210 U/L)

Show code

data1 = filter(dataset, Infused == 1)

km_PFS <- survfit(Surv(Days.to.relapse.death.or.DLC/30, Relapse.or.death) ~ High.LDH, data1)
km_PFS

Call: survfit(formula = Surv(Days.to.relapse.death.or.DLC/30, Relapse.or.death) ~ 
    High.LDH, data = data1)

   1 observation deleted due to missingness 
            n events median 0.95LCL 0.95UCL
High.LDH=0 84     44  19.17    7.20    47.5
High.LDH=1 64     39   3.63    2.83    10.9

Show code

#autoplot(km_Regain)


PFS_plot <- ggsurvplot(km_PFS,
           pval=TRUE,
           conf.int = TRUE,
           risk.table = TRUE, # Add risk table
           fontsize = 6,
           risk.table.col = "strata", # Change risk table color by groups
           tables.height = 0.25,
           tables.theme = theme_cleantable(),
           #surv.median.line = "hv", # Specify median survival
           #ggtheme = theme_bw(), # Change ggplot2 theme
           #palette = c("#E7B800", "#2E9FDF"),
           xlab="Time (months)", ylab = "Progression-free survival",
           xlim = c(0,72), break.x.by = c(12),
           ylim = c(0,1), break.y.by = c(0.25),
           legend.labs = c("LDH <210 U/L", "LDH \u2265210 U/L"),
           legend = "none",
           surv.scale="percent",
           font.main = c(16, "plain", "black"),
           font.x = c(16, "plain", "black"),
           font.y = c(16, "plain", "black"),
           font.caption = c(16, "plain", "black"),
           font.tickslab = c(16, "plain", "black")
)
PFS_plot

Create KM for PFS stratified by HCT-CI status (>=3 vs <3)

Show code

data1 = filter(dataset, Infused == 1) %>%
  mutate(
    High.HCT.CI = ifelse(HCT.CI >=3,1,0) )

km_PFS <- survfit(Surv(Days.to.relapse.death.or.DLC/30, Relapse.or.death) ~ High.HCT.CI, data1 )
km_PFS

Call: survfit(formula = Surv(Days.to.relapse.death.or.DLC/30, Relapse.or.death) ~ 
    High.HCT.CI, data = data1)

                n events median 0.95LCL 0.95UCL
High.HCT.CI=0 100     53   7.43    4.67    24.3
High.HCT.CI=1  49     31   6.07    3.27    41.1

Show code

PFS_plot <- ggsurvplot(km_PFS,
           pval=TRUE,
           conf.int = TRUE,
           risk.table = TRUE, # Add risk table
           fontsize = 6,
           risk.table.col = "strata", # Change risk table color by groups
           tables.height = 0.25,
           tables.theme = theme_cleantable(),
           #surv.median.line = "hv", # Specify median survival
           #ggtheme = theme_bw(), # Change ggplot2 theme
           #palette = c("#E7B800", "#2E9FDF"),
           xlab="Time (months)", ylab = "Progression-free survival",
           xlim = c(0,72), break.x.by = c(12),
           ylim = c(0,1), break.y.by = c(0.25),
           legend.labs = c("HCT <3", "HCT \u22653"),
           legend = "none",
           surv.scale="percent",
           font.main = c(16, "plain", "black"),
           font.x = c(16, "plain", "black"),
           font.y = c(16, "plain", "black"),
           font.caption = c(16, "plain", "black"),
           font.tickslab = c(16, "plain", "black")
)
PFS_plot

Create KM for PFS stratified by bridging response category

Show code

data1 = dataset %>%
  mutate(
    Bridging.response.category = factor(recode(Bridging.response.category, "0" = "No bridging", "1" = "CR/PR", "2" = "SD/PD"), levels = c("No bridging","CR/PR","SD/PD"))
    )

km_PFS <- survfit(Surv(Days.to.relapse.death.or.DLC/30, Relapse.or.death) ~ Bridging.response.category, data1 )
km_PFS

Call: survfit(formula = Surv(Days.to.relapse.death.or.DLC/30, Relapse.or.death) ~ 
    Bridging.response.category, data = data1)

   4 observations deleted due to missingness 
                                        n events median 0.95LCL
Bridging.response.category=No bridging 88     40  39.77   10.33
Bridging.response.category=CR/PR       21     12   7.43    3.67
Bridging.response.category=SD/PD       36     28   3.23    2.23
                                       0.95UCL
Bridging.response.category=No bridging      NA
Bridging.response.category=CR/PR            NA
Bridging.response.category=SD/PD          7.17

Show code

PFS_plot <- ggsurvplot(km_PFS,
           pval=TRUE,
           conf.int = TRUE,
           risk.table = TRUE, # Add risk table
           fontsize = 6,
           risk.table.col = "strata", # Change risk table color by groups
           tables.height = 0.3,
           tables.theme = theme_cleantable(),
           #surv.median.line = "hv", # Specify median survival
           #ggtheme = theme_bw(), # Change ggplot2 theme
           #palette = c("#E7B800", "#2E9FDF"),
           xlab="Time (months)", ylab = "Progression-free survival",
           xlim = c(0,72), break.x.by = c(12),
           ylim = c(0,1), break.y.by = c(0.25),
           legend.labs = c("No bridging","CR/PR","SD/PD"),
           legend = "none",
           surv.scale="percent",
           font.main = c(16, "plain", "black"),
           font.x = c(16, "plain", "black"),
           font.y = c(16, "plain", "black"),
           font.caption = c(16, "plain", "black"),
           font.tickslab = c(16, "plain", "black")
)
PFS_plot

WEIGHTED KM PLOTS

Create weighted KM for OS

Show code

data1 = IPTW.data

km_OS <- survfit(Surv(Days.to.DLC/30, Death) ~ CAR.T.Product.Type, data1, weights=weights)
km_OS

Call: survfit(formula = Surv(Days.to.DLC/30, Death) ~ CAR.T.Product.Type, 
    data = data1, weights = weights)

                            records    n events median 0.95LCL
CAR.T.Product.Type=Breyanzi      50 40.6   17.3   15.6   11.90
CAR.T.Product.Type=Yescarta      99 86.1   40.3   16.3    8.97
                            0.95UCL
CAR.T.Product.Type=Breyanzi      NA
CAR.T.Product.Type=Yescarta      NA

Show code

med_OS <- surv_median(km_OS)

p.logrank <- survdiff(Surv(Days.to.DLC/30, Death) ~ CAR.T.Product.Type, data1)$pvalue


oneYr <- summary(km_OS,times=c(12))
oneYr

Call: survfit(formula = Surv(Days.to.DLC/30, Death) ~ CAR.T.Product.Type, 
    data = data1, weights = weights)

                CAR.T.Product.Type=Breyanzi 
        time       n.risk      n.event     survival      std.err 
     12.0000      16.0336      13.7107       0.6008       0.0883 
lower 95% CI upper 95% CI 
      0.4504       0.8013 

                CAR.T.Product.Type=Yescarta 
        time       n.risk      n.event     survival      std.err 
      12.000       36.110       30.479        0.555        0.061 
lower 95% CI upper 95% CI 
       0.447        0.689 

Show code

oneYr$surv[1]- oneYr$surv[2] # Difference in survival at 1 yr for Breyanzi vs Yescarta

[1] 0.04569224

Show code

print("Difference in survival: ")

[1] "Difference in survival: "

Show code

diffSE <- sqrt(oneYr$std.err[2]^2 + oneYr$std.err[1]^2) # calculate the standard error

#Calculate the 95% CI for the difference
print("95% CI: ")

[1] "95% CI: "

Show code

oneYr$surv[1] - oneYr$surv[2] - 1.96 * diffSE

[1] -0.1646864

Show code

oneYr$surv[1] - oneYr$surv[2] + 1.96 * diffSE

[1] 0.2560708

Show code

zStat <- (oneYr$surv[1] - oneYr$surv[2])/diffSE # calculate the c-test statistic

print("p-value: ")

[1] "p-value: "

Show code

p = 2 * pnorm(abs(zStat), lower.tail = FALSE) # calculate a two-sided p-value


#autoplot(km_Regain)
OS_plot <- ggsurvplot(km_OS,
           pval=FALSE,
           conf.int = TRUE,
           risk.table = TRUE, # Add risk table
           fontsize = 6,
           risk.table.col = "strata", # Change risk table color by groups
           tables.height = 0.25,
           tables.theme = theme_cleantable(),
           xlab="Time (months)", ylab = "Overall survival",
           xlim = c(0,36), break.x.by = c(180/30),
           ylim = c(0,1), break.y.by = c(0.25),
           legend.labs = c("Liso-cel","Axi-cel"),
           legend = "none",
           surv.scale="percent",
           font.main = c(16, "plain", "black"),
           font.x = c(16, "plain", "black"),
           font.y = c(16, "plain", "black"),
           font.caption = c(16, "plain", "black"),
           font.tickslab = c(16, "plain", "black")
) 

OS_plot$plot <- OS_plot$plot + 
  geom_vline(xintercept = 12, linetype = "dashed", color = "steelblue") +
  geom_hline(yintercept = 0.5, linetype = "dashed", color = "black") +
  annotate("text", x = 12.5, y = 0.83, label = paste("Liso-cel vs axi-cel: 1-year OS\n", signif(oneYr$surv[1]*100, 2), "% vs ", signif(oneYr$surv[2]*100, 2), "%, p =", signif(p,2)), color = "steelblue", hjust = 0, vjust = -0.5, size = 3) +
  annotate("text", x = 13, y = 0, label = paste("Liso-cel vs axi-cel,\nmedian OS:\n", signif(med_OS$median[1], 2), "vs", signif(med_OS$median[2], 2), "months, p =", signif(p.logrank,2)), color = "black", hjust = 0, vjust = -0.5, size = 3, fontface = "bold")


OS_plot

Create weighted KM for PFS, stratified by product

Show code

data1 = IPTW.data

km_RFS <- survfit(Surv(Days.to.relapse.death.or.DLC/30, Relapse.or.death) ~ CAR.T.Product.Type, data = data1, weights=weights )
km_RFS

Call: survfit(formula = Surv(Days.to.relapse.death.or.DLC/30, Relapse.or.death) ~ 
    CAR.T.Product.Type, data = data1, weights = weights)

                            records    n events median 0.95LCL
CAR.T.Product.Type=Breyanzi      50 40.6   23.1   7.43    4.67
CAR.T.Product.Type=Yescarta      99 86.1   46.6  10.33    4.90
                            0.95UCL
CAR.T.Product.Type=Breyanzi      NA
CAR.T.Product.Type=Yescarta    47.5

Show code

med_RFS <- surv_median(km_RFS)

p.logrank <- survdiff(Surv(Days.to.relapse.death.or.DLC/30, Relapse.or.death) ~ CAR.T.Product.Type, data1)$pvalue


oneYr <- summary(km_RFS,times=c(12))
oneYr

Call: survfit(formula = Surv(Days.to.relapse.death.or.DLC/30, Relapse.or.death) ~ 
    CAR.T.Product.Type, data = data1, weights = weights)

                CAR.T.Product.Type=Breyanzi 
        time       n.risk      n.event     survival      std.err 
     12.0000       9.6155      20.9003       0.3951       0.0873 
lower 95% CI upper 95% CI 
      0.2562       0.6092 

                CAR.T.Product.Type=Yescarta 
        time       n.risk      n.event     survival      std.err 
     12.0000      30.4878      39.6663       0.4697       0.0597 
lower 95% CI upper 95% CI 
      0.3662       0.6025 

Show code

oneYr$surv[1]- oneYr$surv[2] # Difference in survival at 1 yr for Breyanzi vs Yescarta

[1] -0.07463201

Show code

print("Difference in survival: ")

[1] "Difference in survival: "

Show code

diffSE <- sqrt(oneYr$std.err[2]^2 + oneYr$std.err[1]^2) # calculate the standard error

#Calculate the 95% CI for the difference
print("95% CI: ")

[1] "95% CI: "

Show code

oneYr$surv[1] - oneYr$surv[2] - 1.96 * diffSE

[1] -0.281858

Show code

oneYr$surv[1] - oneYr$surv[2] + 1.96 * diffSE

[1] 0.132594

Show code

zStat <- (oneYr$surv[1] - oneYr$surv[2])/diffSE # calculate the c-test statistic

print("p-value: ")

[1] "p-value: "

Show code

p = 2 * pnorm(abs(zStat), lower.tail = FALSE) # calculate a two-sided p-value


#autoplot(km_Regain)
RFS_plot <- ggsurvplot(km_RFS,
           pval=FALSE,
           conf.int = TRUE,
           risk.table = TRUE, # Add risk table
           fontsize = 6,
           risk.table.col = "strata", # Change risk table color by groups
           tables.height = 0.25,
           tables.theme = theme_cleantable(),
           xlab="Time (months)", ylab = "Progression-free survival",
           xlim = c(0,36), break.x.by = c(180/30),
           ylim = c(0,1), break.y.by = c(0.25),
           legend.labs = c("Liso-cel","Axi-cel"),
           legend = "none",
           surv.scale="percent",
           font.main = c(16, "plain", "black"),
           font.x = c(16, "plain", "black"),
           font.y = c(16, "plain", "black"),
           font.caption = c(16, "plain", "black"),
           font.tickslab = c(16, "plain", "black")
) 

RFS_plot$plot <- RFS_plot$plot + 
  geom_vline(xintercept = 12, linetype = "dashed", color = "steelblue") +
  geom_hline(yintercept = 0.5, linetype = "dashed", color = "black") +
  annotate("text", x = 12.5, y = 0.83, label = paste("Liso-cel vs axi-cel: 1-year PFS\n", signif(oneYr$surv[1]*100, 2), "% vs ", signif(oneYr$surv[2]*100, 2), "%, p =", signif(p,2)), color = "steelblue", hjust = 0, vjust = -0.5, size = 3) +
  annotate("text", x = 13, y = .55, label = paste("Liso-cel vs axi-cel,\nmedian PFS:\n", signif(med_RFS$median[1], 2), "vs", signif(med_RFS$median[2], 2), "months, p =", signif(p.logrank,2)), color = "black", hjust = 0, vjust = -0.5, size = 3, fontface = "bold")


RFS_plot

REGRESSION ANALYSIS: UNWEIGHTED AND IPTW

PFS/OS

Cox Model - PFS

Show code

theme_gtsummary_compact()
preds <- c("Product", "LDH pre-LD", "ALC pre-LD","Largest lesion", "Age", "Male sex", "Bridging category")

data0 <- dataset %>%
  mutate(
    "Male sex" = ifelse(Sex == "M",1,0),
    Product = factor(recode(CAR.T.Product.Type, "Yescarta" = "Axi-cel","Breyanzi" = "Liso-cel"), levels = c("Liso-cel","Axi-cel") ),
    "ALC pre-LD" = ALC.preLD,
    "LDH pre-LD" = LDH.preLD/1000,
    "Largest lesion" = Largest.lesion,
    "Bridging category" = factor(recode(Bridging.response.category, "0" = "No bridging", "1" = "CR/PR", "2" = "SD/PD"), levels = c("No bridging","CR/PR","SD/PD"))
  )

data1 <- IPTW.data %>%
  mutate(
    "Male sex" = ifelse(Sex == "M",1,0),
    Product = factor(recode(CAR.T.Product.Type, "Yescarta" = "Axi-cel","Breyanzi" = "Liso-cel"), levels = c("Liso-cel","Axi-cel") ),
    "ALC pre-LD" = ALC.preLD,
    "LDH pre-LD" = LDH.preLD/1000,
    "Largest lesion" = Largest.lesion,
    "Bridging category" = factor(recode(Bridging.response.category, "0" = "No bridging", "1" = "CR/PR", "2" = "SD/PD"), levels = c("No bridging","CR/PR","SD/PD")),
  )

data2 <- m.data %>%
  mutate(
    "Male sex" = ifelse(Sex == "M",1,0),
    Product = factor(recode(CAR.T.Product.Type, "Yescarta" = "Axi-cel","Breyanzi" = "Liso-cel"), levels = c("Liso-cel","Axi-cel") ),
    "ALC pre-LD" = ALC.preLD,
    "LDH pre-LD" = LDH.preLD/1000,
    "Largest lesion" = Largest.lesion,
    "Bridging category" = factor(recode(Bridging.response.category, "0" = "No bridging", "1" = "CR/PR", "2" = "SD/PD"), levels = c("No bridging","CR/PR","SD/PD")),
  )

uv_tab <- tbl_uvregression(
  data0[c(preds)],
  method = coxph,
  y = Surv(data0$Days.to.relapse.death.or.DLC, data0$Relapse.or.death),
  exponentiate = TRUE
)

mv_tab<-coxph( Surv(Days.to.relapse.death.or.DLC, Relapse.or.death) ~ Product + `LDH pre-LD`+  `ALC pre-LD` + `Largest lesion` + Age+ `Male sex` + `Bridging category`, data0) %>% tbl_regression(exponentiate=TRUE)


uv_tab_IPTW <- tbl_uvregression(
  data1[c(preds)],
  method = coxph,
  y = Surv(data1$Days.to.relapse.death.or.DLC, data1$Relapse.or.death),
  exponentiate = TRUE,
  method.args = list(weights = data1$weights),
)

mv_tab_IPTW<-coxph( Surv(Days.to.relapse.death.or.DLC, Relapse.or.death) ~ Product + `LDH pre-LD`+  `ALC pre-LD` + `Largest lesion` + Age + `Male sex` + `Bridging category`, data1, weights=weights) %>%tbl_regression( exponentiate=TRUE)

uv_tab_PSM <- tbl_uvregression(
  data2[c(preds)],
  method = coxph,
  y = Surv(data2$Days.to.relapse.death.or.DLC, data2$Relapse.or.death),
  exponentiate = TRUE,
  method.args = list(weights = data2$weights),
)

mv_tab_PSM<-coxph( Surv(Days.to.relapse.death.or.DLC, Relapse.or.death) ~ Product + `LDH pre-LD`+  `ALC pre-LD` + `Largest lesion` + Age + `Male sex` + `Bridging category`, data2, weights=weights) %>%tbl_regression( exponentiate=TRUE)


row1 <- tbl_merge(list(uv_tab, mv_tab), tab_spanner = c("**Univariate**", "**Multivariable**"))
row2 <- tbl_merge(list(uv_tab_IPTW, mv_tab_IPTW))
row3 <- tbl_merge(list(uv_tab_PSM, mv_tab_PSM))
tbl_stack(list(row1, row2, row3), group_header = c("UNWEIGHTED ANALYSIS", "IPTW ANALYSIS", "PSM ANALYSIS"), quiet=TRUE)

Characteristic	Univariate				Multivariable
	N	HR1	95% CI1	p-value	HR1	95% CI1	p-value
UNWEIGHTED ANALYSIS
Product	149
Liso-cel		—	—		—	—
Axi-cel		0.85	0.54, 1.34	0.5	0.93	0.56, 1.54	0.8
LDH pre-LD	148	1.14	1.01, 1.28	0.037	1.10	0.96, 1.26	0.2
ALC pre-LD	149	1.11	0.96, 1.29	0.2	1.11	0.99, 1.25	0.069
Largest lesion	149	1.08	1.02, 1.14	0.011	1.05	0.98, 1.12	0.2
Age	149	1.0	0.98, 1.01	0.5	1.00	0.98, 1.02	0.7
Male sex	149	1.05	0.67, 1.64	0.8	1.28	0.78, 2.09	0.3
Bridging category	145
No bridging		—	—		—	—
CR/PR		1.58	0.83, 3.04	0.2	1.69	0.86, 3.32	0.13
SD/PD		2.68	1.63, 4.39	<0.001	2.31	1.29, 4.13	0.005
IPTW ANALYSIS
Product	126
Liso-cel		—	—		—	—
Axi-cel		0.90	0.56, 1.46	0.7	1.07	0.64, 1.79	0.8
LDH pre-LD	125	1.14	1.08, 1.21	<0.001	1.11	1.04, 1.18	0.001
ALC pre-LD	126	1.12	0.96, 1.31	0.15	1.13	1.05, 1.21	0.001
Largest lesion	126	1.05	0.99, 1.12	0.11	1.01	0.94, 1.09	0.7
Age	126	1.00	0.98, 1.02	0.8	1.00	0.98, 1.02	0.8
Male sex	126	1.03	0.63, 1.69	0.9	1.29	0.74, 2.25	0.4
Bridging category	123
No bridging		—	—		—	—
CR/PR		1.61	0.69, 3.74	0.3	1.71	0.67, 4.33	0.3
SD/PD		2.62	1.60, 4.29	<0.001	2.60	1.44, 4.71	0.002
PSM ANALYSIS
Product	149
Liso-cel		—	—		—	—
Axi-cel		0.79	0.45, 1.39	0.4	0.87	0.50, 1.52	0.6
LDH pre-LD	148	1.23	1.09, 1.39	<0.001	1.19	1.06, 1.33	0.003
ALC pre-LD	149	1.16	1.07, 1.26	<0.001	1.15	1.09, 1.22	<0.001
Largest lesion	149	1.05	0.98, 1.13	0.2	0.99	0.92, 1.06	0.8
Age	149	0.99	0.97, 1.01	0.5	1.01	0.98, 1.03	0.6
Male sex	149	1.46	0.80, 2.67	0.2	1.71	0.99, 2.94	0.052
Bridging category	144
No bridging		—	—		—	—
CR/PR		1.04	0.32, 3.37	>0.9	1.02	0.31, 3.35	>0.9
SD/PD		2.77	1.59, 4.81	<0.001	2.86	1.62, 5.07	<0.001
1 HR = Hazard Ratio, CI = Confidence Interval

Cox Model - PFS forest plots

Show code

## Forest plot for MV
df_fp <- mv_tab$table_body %>%
  select(header_row, label, N_obs, N_event, estimate, conf.low, conf.high, p.value, ) %>%
  mutate(
    label = ifelse(header_row == TRUE | is.na(header_row),as.character(label), paste0("     ", as.character(label)) ),
    p.value = signif(p.value, 2),
    `HR (95% CI)` = ifelse(
      is.na(estimate),"",paste0(round(estimate, 2), " (", round(conf.low,2), " to ",round(conf.high,2), ") ")
    )
  ) %>%
  rename(Characteristic = label,
         N = N_obs,
         Events = N_event,
         `p value` = p.value)
forester(
  left_side_data = df_fp[2],
  right_side_data = df_fp[9:8],
  estimate = df_fp$estimate,
  ci_low = df_fp$conf.low,
  ci_high = df_fp$conf.high,
  estimate_col_name = "HR (95% CI)",
  xlim = c(0.1, 10),
  null_line_at = 1,
  arrows = TRUE,
  x_scale_linear = FALSE,
  arrow_labels = c("Better", "Worse"),
  font_family = "sans",
  display = TRUE,
  nudge_x = .4,
  nudge_height = .06,
  stripe_colour = "#F5F5F5",
  file_path = "/Users/aportugu/Desktop/CD19 project/ASH abstract/forest/PFS_mv_plot.png"
)

knitr::include_graphics("/Users/aportugu/Desktop/CD19 project/ASH abstract/forest/PFS_mv_plot.png")

Show code

## Forest plot for MV IPTW
df_fp <- mv_tab_IPTW$table_body %>%
  select(header_row, label, N_obs, N_event, estimate, conf.low, conf.high, p.value, ) %>%
  mutate(
    label = ifelse(header_row == TRUE | is.na(header_row),as.character(label), paste0("     ", as.character(label)) ),
    p.value = signif(p.value, 2),
    `HR (95% CI)` = ifelse(
      is.na(estimate),"",paste0(round(estimate, 2), " (", round(conf.low,2), " to ",round(conf.high,2), ") ")
    )
  ) %>%
  rename(Characteristic = label,
         N = N_obs,
         Events = N_event,
         `p value` = p.value)
forester(
  left_side_data = df_fp[2],
  right_side_data = df_fp[9:8],
  estimate = df_fp$estimate,
  ci_low = df_fp$conf.low,
  ci_high = df_fp$conf.high,
  estimate_col_name = "HR (95% CI)",
  xlim = c(0.1, 10),
  null_line_at = 1,
  arrows = TRUE,
  x_scale_linear = FALSE,
  arrow_labels = c("Better", "Worse"),
  font_family = "sans",
  display = TRUE,
  nudge_x = .4,
  nudge_height = .06,
  stripe_colour = "#F5F5F5",
  file_path = "/Users/aportugu/Desktop/CD19 project/ASH abstract/forest/PFS_mv_plot_IPTW.png"
)

knitr::include_graphics("/Users/aportugu/Desktop/CD19 project/ASH abstract/forest/PFS_mv_plot_IPTW.png")

Show code

## Forest plot for MV PSM
df_fp <- mv_tab_PSM$table_body %>%
  select(header_row, label, N_obs, N_event, estimate, conf.low, conf.high, p.value, ) %>%
  mutate(
    label = ifelse(header_row == TRUE | is.na(header_row),as.character(label), paste0("     ", as.character(label)) ),
    p.value = signif(p.value, 2),
    `HR (95% CI)` = ifelse(
      is.na(estimate),"",paste0(round(estimate, 2), " (", round(conf.low,2), " to ",round(conf.high,2), ") ")
    )
  ) %>%
  rename(Characteristic = label,
         N = N_obs,
         Events = N_event,
         `p value` = p.value)
forester(
  left_side_data = df_fp[2],
  right_side_data = df_fp[9:8],
  estimate = df_fp$estimate,
  ci_low = df_fp$conf.low,
  ci_high = df_fp$conf.high,
  estimate_col_name = "HR (95% CI)",
  xlim = c(0.1, 10),
  null_line_at = 1,
  arrows = TRUE,
  x_scale_linear = FALSE,
  arrow_labels = c("Better", "Worse"),
  font_family = "sans",
  display = TRUE,
  nudge_x = .4,
  nudge_height = .06,
  stripe_colour = "#F5F5F5",
  file_path = "/Users/aportugu/Desktop/CD19 project/ASH abstract/forest/PFS_mv_plot_PSM.png"
)

knitr::include_graphics("/Users/aportugu/Desktop/CD19 project/ASH abstract/forest/PFS_mv_plot_PSM.png")

Cox Model - OS

Show code

theme_gtsummary_compact()
preds <- c("Product","LDH pre-LD", "ALC pre-LD","Largest lesion", "Age", "Male sex", "Bridging category")

data0 <- dataset %>%
  mutate(
    "Male sex" = ifelse(Sex == "M",1,0),
    Product = factor(recode(CAR.T.Product.Type, "Yescarta" = "Axi-cel","Breyanzi" = "Liso-cel"), levels = c("Liso-cel","Axi-cel") ),
    "ALC pre-LD" = ALC.preLD,
    "LDH pre-LD" = LDH.preLD/1000,
    "Largest lesion" = Largest.lesion,
    "Bridging category" = factor(recode(Bridging.response.category, "0" = "No bridging", "1" = "CR/PR", "2" = "SD/PD"), levels = c("No bridging","CR/PR","SD/PD"))
  )

data1 <- IPTW.data %>%
  mutate(
    "Male sex" = ifelse(Sex == "M",1,0),
    Product = factor(recode(CAR.T.Product.Type, "Yescarta" = "Axi-cel","Breyanzi" = "Liso-cel"), levels = c("Liso-cel","Axi-cel") ),
    "ALC pre-LD" = ALC.preLD,
    "LDH pre-LD" = LDH.preLD/1000,
    "Largest lesion" = Largest.lesion,
    "Bridging category" = factor(recode(Bridging.response.category, "0" = "No bridging", "1" = "CR/PR", "2" = "SD/PD"), levels = c("No bridging","CR/PR","SD/PD"))
  )

data2 <- m.data %>%
  mutate(
    "Male sex" = ifelse(Sex == "M",1,0),
    Product = factor(recode(CAR.T.Product.Type, "Yescarta" = "Axi-cel","Breyanzi" = "Liso-cel"), levels = c("Liso-cel","Axi-cel") ),
    "ALC pre-LD" = ALC.preLD,
    "LDH pre-LD" = LDH.preLD/1000,
    "Largest lesion" = Largest.lesion,
    "Bridging category" = factor(recode(Bridging.response.category, "0" = "No bridging", "1" = "CR/PR", "2" = "SD/PD"), levels = c("No bridging","CR/PR","SD/PD"))
  )

uv_tab <- tbl_uvregression(
  data0[c(preds)],
  method = coxph,
  y = Surv(data0$Days.to.DLC, data0$Death),
  exponentiate = TRUE
)

mv_tab<-coxph( Surv(Days.to.DLC, Death) ~ Product +`LDH pre-LD` +  `ALC pre-LD` +  `Largest lesion` + Age + `Male sex` + `Bridging category`, data0) %>% tbl_regression(exponentiate=TRUE)


uv_tab_IPTW <- tbl_uvregression(
  data1[c(preds)],
  method = coxph,
  y = Surv(data1$Days.to.DLC, data1$Death),
  exponentiate = TRUE,
  method.args = list(weights = data1$weights),
)

mv_tab_IPTW<-coxph( Surv(Days.to.DLC, Death) ~ Product +`LDH pre-LD` +  `ALC pre-LD` +  `Largest lesion` + Age + `Male sex` + `Bridging category`, data1, weights=weights) %>% tbl_regression(exponentiate=TRUE)

uv_tab_PSM <- tbl_uvregression(
  data2[c(preds)],
  method = coxph,
  y = Surv(data2$Days.to.DLC, data2$Death),
  exponentiate = TRUE,
  method.args = list(weights = data2$weights),
)

mv_tab_PSM<-coxph( Surv(Days.to.DLC, Death) ~ Product +`LDH pre-LD` +  `ALC pre-LD` +  `Largest lesion` + Age + `Male sex` + `Bridging category`, data2, weights=weights) %>% tbl_regression(exponentiate=TRUE)



row1 <- tbl_merge(list(uv_tab, mv_tab), tab_spanner = c("**Univariate**", "**Multivariable**"))
row2 <- tbl_merge(list(uv_tab_IPTW, mv_tab_IPTW))
row3 <- tbl_merge(list(uv_tab_PSM, mv_tab_PSM))
tbl_stack(list(row1, row2, row3), group_header = c("UNWEIGHTED ANALYSIS", "IPTW ANALYSIS", "PSM ANALYSIS"), quiet=TRUE)

Characteristic	Univariate				Multivariable
	N	HR1	95% CI1	p-value	HR1	95% CI1	p-value
UNWEIGHTED ANALYSIS
Product	149
Liso-cel		—	—		—	—
Axi-cel		1.19	0.69, 2.03	0.5	1.17	0.65, 2.13	0.6
LDH pre-LD	148	1.37	1.19, 1.58	<0.001	1.37	1.16, 1.61	<0.001
ALC pre-LD	149	1.23	1.04, 1.47	0.017	1.27	1.06, 1.52	0.010
Largest lesion	149	1.14	1.07, 1.21	<0.001	1.11	1.04, 1.19	0.002
Age	149	1.00	0.98, 1.02	0.7	1.00	0.98, 1.02	0.8
Male sex	149	1.12	0.66, 1.88	0.7	1.36	0.77, 2.39	0.3
Bridging category	145
No bridging		—	—		—	—
CR/PR		1.52	0.72, 3.21	0.3	1.85	0.85, 4.02	0.12
SD/PD		2.77	1.60, 4.78	<0.001	2.07	1.09, 3.93	0.026
IPTW ANALYSIS
Product	126
Liso-cel		—	—		—	—
Axi-cel		1.10	0.62, 1.95	0.7	1.17	0.63, 2.17	0.6
LDH pre-LD	125	1.36	1.22, 1.51	<0.001	1.34	1.20, 1.50	<0.001
ALC pre-LD	126	1.24	1.07, 1.44	0.005	1.28	1.17, 1.41	<0.001
Largest lesion	126	1.13	1.07, 1.20	<0.001	1.10	1.03, 1.18	0.003
Age	126	1.00	0.98, 1.02	0.8	1.00	0.97, 1.02	0.8
Male sex	126	1.06	0.60, 1.85	0.8	1.38	0.73, 2.61	0.3
Bridging category	123
No bridging		—	—		—	—
CR/PR		1.51	0.64, 3.53	0.3	1.74	0.68, 4.45	0.3
SD/PD		2.96	1.66, 5.27	<0.001	2.35	1.21, 4.59	0.012
PSM ANALYSIS
Product	149
Liso-cel		—	—		—	—
Axi-cel		1.06	0.56, 1.98	0.9	1.14	0.59, 2.21	0.7
LDH pre-LD	148	1.49	1.30, 1.72	<0.001	1.45	1.24, 1.71	<0.001
ALC pre-LD	149	1.33	1.21, 1.45	<0.001	1.38	1.12, 1.70	0.003
Largest lesion	149	1.14	1.07, 1.22	<0.001	1.09	1.01, 1.17	0.028
Age	149	1.00	0.96, 1.03	0.9	1.00	0.97, 1.05	0.8
Male sex	149	1.32	0.68, 2.59	0.4	1.26	0.66, 2.40	0.5
Bridging category	144
No bridging		—	—		—	—
CR/PR		1.29	0.40, 4.12	0.7	1.41	0.42, 4.72	0.6
SD/PD		3.15	1.56, 6.33	0.001	2.68	1.31, 5.48	0.007
1 HR = Hazard Ratio, CI = Confidence Interval

Cox Model - OS forest plots

Show code

## Forest plot for MV
df_fp <- mv_tab$table_body %>%
  select(header_row, label, N_obs, N_event, estimate, conf.low, conf.high, p.value, ) %>%
  mutate(
    label = ifelse(header_row == TRUE | is.na(header_row),as.character(label), paste0("     ", as.character(label)) ),
    p.value = signif(p.value, 2),
    `HR (95% CI)` = ifelse(
      is.na(estimate),"",paste0(round(estimate, 2), " (", round(conf.low,2), " to ",round(conf.high,2), ") ")
    )
  ) %>%
  rename(Characteristic = label,
         N = N_obs,
         Events = N_event,
         `p value` = p.value)
forester(
  left_side_data = df_fp[2],
  right_side_data = df_fp[9:8],
  estimate = df_fp$estimate,
  ci_low = df_fp$conf.low,
  ci_high = df_fp$conf.high,
  estimate_col_name = "HR (95% CI)",
  xlim = c(0.1, 10),
  null_line_at = 1,
  arrows = TRUE,
  x_scale_linear = FALSE,
  arrow_labels = c("Better", "Worse"),
  font_family = "sans",
  display = TRUE,
  nudge_x = .4,
  nudge_height = .06,
  stripe_colour = "#F5F5F5",
  file_path = "/Users/aportugu/Desktop/CD19 project/ASH abstract/forest/OS_mv_plot.png"
)

knitr::include_graphics("/Users/aportugu/Desktop/CD19 project/ASH abstract/forest/OS_mv_plot.png")

Show code

## Forest plot for MV IPTW
df_fp <- mv_tab_IPTW$table_body %>%
  select(header_row, label, N_obs, N_event, estimate, conf.low, conf.high, p.value, ) %>%
  mutate(
    label = ifelse(header_row == TRUE | is.na(header_row),as.character(label), paste0("     ", as.character(label)) ),
    p.value = signif(p.value, 2),
    `HR (95% CI)` = ifelse(
      is.na(estimate),"",paste0(round(estimate, 2), " (", round(conf.low,2), " to ",round(conf.high,2), ") ")
    )
  ) %>%
  rename(Characteristic = label,
         N = N_obs,
         Events = N_event,
         `p value` = p.value)
forester(
  left_side_data = df_fp[2],
  right_side_data = df_fp[9:8],
  estimate = df_fp$estimate,
  ci_low = df_fp$conf.low,
  ci_high = df_fp$conf.high,
  estimate_col_name = "HR (95% CI)",
  xlim = c(0.1, 10),
  null_line_at = 1,
  arrows = TRUE,
  x_scale_linear = FALSE,
  arrow_labels = c("Better", "Worse"),
  font_family = "sans",
  display = TRUE,
  nudge_x = .4,
  nudge_height = .06,
  stripe_colour = "#F5F5F5",
  file_path = "/Users/aportugu/Desktop/CD19 project/ASH abstract/forest/OS_mv_plot_IPTW.png"
)

knitr::include_graphics("/Users/aportugu/Desktop/CD19 project/ASH abstract/forest/OS_mv_plot_IPTW.png")

Show code

## Forest plot for MV PSM
df_fp <- mv_tab_PSM$table_body %>%
  select(header_row, label, N_obs, N_event, estimate, conf.low, conf.high, p.value, ) %>%
  mutate(
    label = ifelse(header_row == TRUE | is.na(header_row),as.character(label), paste0("     ", as.character(label)) ),
    p.value = signif(p.value, 2),
    `HR (95% CI)` = ifelse(
      is.na(estimate),"",paste0(round(estimate, 2), " (", round(conf.low,2), " to ",round(conf.high,2), ") ")
    )
  ) %>%
  rename(Characteristic = label,
         N = N_obs,
         Events = N_event,
         `p value` = p.value)
forester(
  left_side_data = df_fp[2],
  right_side_data = df_fp[9:8],
  estimate = df_fp$estimate,
  ci_low = df_fp$conf.low,
  ci_high = df_fp$conf.high,
  estimate_col_name = "HR (95% CI)",
  xlim = c(0.1, 10),
  null_line_at = 1,
  arrows = TRUE,
  x_scale_linear = FALSE,
  arrow_labels = c("Better", "Worse"),
  font_family = "sans",
  display = TRUE,
  nudge_x = .4,
  nudge_height = .06,
  stripe_colour = "#F5F5F5",
  file_path = "/Users/aportugu/Desktop/CD19 project/ASH abstract/forest/OS_mv_plot_PSM.png"
)

knitr::include_graphics("/Users/aportugu/Desktop/CD19 project/ASH abstract/forest/OS_mv_plot_PSM.png")

Cox diagnostics

PFS unweighted

Show code

data1 <- dataset %>%
  mutate(
    ferritin.max = log(ferritin.max),
    IL6.max = log(IL6.max),
    LDH.min = log(LDH.min),
    Age = Age,
    `Male sex` = ifelse(Sex == "M",1,0),
    Product = factor(recode(CAR.T.Product.Type, "Yescarta" = "Axi-cel","Breyanzi" = "Liso-cel"), levels = c("Liso-cel","Axi-cel") ),
    `ALC pre-LD` = ALC.preLD,
    `LDH pre-LD` = LDH.preLD/1000,
    `High LDH pre-LD` = ifelse(LDH.preLD >=210,1,0),
    `LDH day 0` = LDH.d0/1000,
    `ALC day 0` = ALC.d0,
    `Largest lesion` = Largest.lesion
  )

res.cox <-coxph( Surv(Days.to.relapse.death.or.DLC, Relapse.or.death) ~ Product + `LDH pre-LD` + `ALC pre-LD` + `Largest lesion` + Age + `Male sex`, data1)

## Testing proportional hazards assumption
test.ph <- cox.zph(res.cox)
test.ph

                  chisq df    p
Product           2.188  1 0.14
`LDH pre-LD`      2.538  1 0.11
`ALC pre-LD`      2.137  1 0.14
`Largest lesion`  1.293  1 0.26
Age               1.365  1 0.24
`Male sex`        0.215  1 0.64
GLOBAL           10.147  6 0.12

PFS IPTW

Show code

data1 <- IPTW.data %>%
  mutate(
    `Male sex` = ifelse(Sex == "M",1,0),
    Product = factor(recode(CAR.T.Product.Type, "Yescarta" = "Axi-cel","Breyanzi" = "Liso-cel"), levels = c("Liso-cel","Axi-cel") ),
    `ALC pre-LD` = ALC.preLD,
    `LDH pre-LD` = LDH.preLD/1000,
    `High LDH pre-LD` = ifelse(LDH.preLD >=210,1,0),
    `LDH day 0` = LDH.d0/1000,
    `ALC day 0` = ALC.d0,
    `Largest lesion` = Largest.lesion
  )

res.cox <-coxph( Surv(Days.to.relapse.death.or.DLC, Relapse.or.death) ~ Product + `LDH pre-LD` + `ALC pre-LD` + `Largest lesion`+ Age + `Male sex`, data1, weights=weights)

## Testing proportional hazards assumption
test.ph <- cox.zph(res.cox)
test.ph

                  chisq df     p
Product           3.459  1 0.063
`LDH pre-LD`      2.039  1 0.153
`ALC pre-LD`      1.558  1 0.212
`Largest lesion`  0.370  1 0.543
Age               2.558  1 0.110
`Male sex`        0.179  1 0.672
GLOBAL           10.079  6 0.121

PFS PSM

Show code

data1 <- m.data %>%
  mutate(
    `Male sex` = ifelse(Sex == "M",1,0),
    Product = factor(recode(CAR.T.Product.Type, "Yescarta" = "Axi-cel","Breyanzi" = "Liso-cel"), levels = c("Liso-cel","Axi-cel") ),
    `ALC pre-LD` = ALC.preLD,
    `LDH pre-LD` = LDH.preLD/1000,
    `High LDH pre-LD` = ifelse(LDH.preLD >=210,1,0),
    `LDH day 0` = LDH.d0/1000,
    `ALC day 0` = ALC.d0,
    `Largest lesion` = Largest.lesion
  )

res.cox <-coxph( Surv(Days.to.relapse.death.or.DLC, Relapse.or.death) ~ Product + `LDH pre-LD` + `ALC pre-LD` + `Largest lesion`+ Age + `Male sex`, data1, weights=weights)

## Testing proportional hazards assumption
test.ph <- cox.zph(res.cox)
test.ph

                 chisq df    p
Product          1.423  1 0.23
`LDH pre-LD`     0.991  1 0.32
`ALC pre-LD`     0.293  1 0.59
`Largest lesion` 0.338  1 0.56
Age              0.134  1 0.71
`Male sex`       1.066  1 0.30
GLOBAL           4.275  6 0.64

OS unweighted

Show code

data1 <- dataset %>%
  mutate(
    ferritin.max = log(ferritin.max),
    IL6.max = log(IL6.max),
    LDH.min = log(LDH.min),
    Age = Age,
    `Male sex` = ifelse(Sex == "M",1,0),
    Product = factor(recode(CAR.T.Product.Type, "Yescarta" = "Axi-cel","Breyanzi" = "Liso-cel"), levels = c("Liso-cel","Axi-cel") ),
    `ALC pre-LD` = ALC.preLD,
    `LDH pre-LD` = LDH.preLD/1000,
    `High LDH pre-LD` = ifelse(LDH.preLD >=210,1,0),
    `LDH day 0` = LDH.d0/1000,
    `ALC day 0` = ALC.d0,
    `Largest lesion` = Largest.lesion
  )

res.cox <-coxph( Surv(Days.to.DLC, Death) ~ Product + `LDH pre-LD` + `ALC pre-LD` + `Largest lesion`+ Age + `Male sex`, data1)

## Testing proportional hazards assumption
test.ph <- cox.zph(res.cox)
test.ph

                   chisq df     p
Product          5.85230  1 0.016
`LDH pre-LD`     0.00569  1 0.940
`ALC pre-LD`     1.17243  1 0.279
`Largest lesion` 2.27721  1 0.131
Age              1.69047  1 0.194
`Male sex`       0.01424  1 0.905
GLOBAL           9.93558  6 0.127

OS IPTW

Show code

data1 <- IPTW.data %>%
  mutate(
    ferritin.max = log(ferritin.max),
    IL6.max = log(IL6.max),
    LDH.min = log(LDH.min),
    Age = Age,
    `Male sex` = ifelse(Sex == "M",1,0),
    Product = factor(recode(CAR.T.Product.Type, "Yescarta" = "Axi-cel","Breyanzi" = "Liso-cel"), levels = c("Liso-cel","Axi-cel") ),
    `ALC pre-LD` = ALC.preLD,
    `LDH pre-LD` = LDH.preLD/1000,
    `High LDH pre-LD` = ifelse(LDH.preLD >=210,1,0),
    `LDH day 0` = LDH.d0/1000,
    `ALC day 0` = ALC.d0,
    `Largest lesion` = Largest.lesion
  )

res.cox <-coxph( Surv(Days.to.DLC, Death) ~ Product + `LDH pre-LD` + `ALC pre-LD` + `Largest lesion`+ Age + `Male sex`, data1, weights=weights)

## Testing proportional hazards assumption
test.ph <- cox.zph(res.cox)
test.ph

                  chisq df    p
Product          2.0630  1 0.15
`LDH pre-LD`     0.0402  1 0.84
`ALC pre-LD`     0.9209  1 0.34
`Largest lesion` 2.6437  1 0.10
Age              1.9744  1 0.16
`Male sex`       0.0519  1 0.82
GLOBAL           8.2084  6 0.22

OS PSM

Show code

data1 <- m.data %>%
  mutate(
    ferritin.max = log(ferritin.max),
    IL6.max = log(IL6.max),
    LDH.min = log(LDH.min),
    Age = Age,
    `Male sex` = ifelse(Sex == "M",1,0),
    Product = factor(recode(CAR.T.Product.Type, "Yescarta" = "Axi-cel","Breyanzi" = "Liso-cel"), levels = c("Liso-cel","Axi-cel") ),
    `ALC pre-LD` = ALC.preLD,
    `LDH pre-LD` = LDH.preLD/1000,
    `High LDH pre-LD` = ifelse(LDH.preLD >=210,1,0),
    `LDH day 0` = LDH.d0/1000,
    `ALC day 0` = ALC.d0,
    `Largest lesion` = Largest.lesion
  )

res.cox <-coxph( Surv(Days.to.DLC, Death) ~ Product + `LDH pre-LD` + `ALC pre-LD` + `Largest lesion`+ Age + `Male sex`, data1, weights=weights)

## Testing proportional hazards assumption
test.ph <- cox.zph(res.cox)
test.ph

                    chisq df     p
Product          2.54e+00  1 0.111
`LDH pre-LD`     5.45e-01  1 0.460
`ALC pre-LD`     2.65e-06  1 0.999
`Largest lesion` 1.24e+00  1 0.266
Age              5.66e+00  1 0.017
`Male sex`       2.56e-01  1 0.613
GLOBAL           1.14e+01  6 0.076

Response

Logistic Model - CR

Show code

theme_gtsummary_compact()
preds <- c("Best.CR", "Product", "LDH pre-LD","ALC pre-LD", "Largest lesion", "Age", "Male sex", "Bridging category") 

data0 <- dataset %>%
  mutate(
    Product = factor(recode(CAR.T.Product.Type, "Yescarta" = "Axi-cel","Breyanzi" = "Liso-cel"), levels = c("Liso-cel","Axi-cel") ),
    "Largest lesion" = Largest.lesion,
    "ALC pre-LD" = ALC.preLD,
    "LDH pre-LD" = LDH.preLD/1000,
    "Male sex" = ifelse(Sex == "M",1,0),
    "Bridging category" = factor(recode(Bridging.response.category, "0" = "No bridging", "1" = "CR/PR", "2" = "SD/PD"), levels = c("No bridging","CR/PR","SD/PD"))
  )

data1 <- IPTW.data %>%
  mutate(
    Product = factor(recode(CAR.T.Product.Type, "Yescarta" = "Axi-cel","Breyanzi" = "Liso-cel"), levels = c("Liso-cel","Axi-cel") ),
    "Largest lesion" = Largest.lesion,
    "ALC pre-LD" = ALC.preLD,
    "LDH pre-LD" = LDH.preLD/1000,
    "Male sex" = ifelse(Sex == "M",1,0),
    "Bridging category" = factor(recode(Bridging.response.category, "0" = "No bridging", "1" = "CR/PR", "2" = "SD/PD"), levels = c("No bridging","CR/PR","SD/PD"))
  )

data2 <- m.data %>%
  mutate(
    Product = factor(recode(CAR.T.Product.Type, "Yescarta" = "Axi-cel","Breyanzi" = "Liso-cel"), levels = c("Liso-cel","Axi-cel") ),
    "Largest lesion" = Largest.lesion,
    "ALC pre-LD" = ALC.preLD,
    "LDH pre-LD" = LDH.preLD/1000,
    "Male sex" = ifelse(Sex == "M",1,0),
    "Bridging category" = factor(recode(Bridging.response.category, "0" = "No bridging", "1" = "CR/PR", "2" = "SD/PD"), levels = c("No bridging","CR/PR","SD/PD"))
  )

uv_tab <- tbl_uvregression(
  data0[c(preds)],
  method = glm,
  y = Best.CR,
  method.args = list(family = binomial),
  exponentiate = TRUE
)


mv_tab<-glm( Best.CR ~  `LDH pre-LD` + Product + `Largest lesion` + `ALC pre-LD` + `Age` + `Male sex` + `Bridging category`, data0, family = binomial ) %>%
  tbl_regression(exponentiate=TRUE)

uv_tab_IPTW <- tbl_uvregression(
  data1[c(preds)],
  method = glm,
  y = Best.CR,
  method.args = list(family = binomial, weights = data1$weights),
  exponentiate = TRUE
)

mv_tab_IPTW<-glm( Best.CR ~  `LDH pre-LD` +Product + `Largest lesion` + `ALC pre-LD` +  `Age` + `Male sex` + `Bridging category`, data1, weights=weights, family = binomial ) %>% 
  tbl_regression(exponentiate=TRUE)


uv_tab_PSM <- tbl_uvregression(
  data2[c(preds)],
  method = glm,
  y = Best.CR,
  method.args = list(family = binomial, weights = data2$weights),
  exponentiate = TRUE
)

mv_tab_PSM<-glm( Best.CR ~  `LDH pre-LD` +Product + `Largest lesion` + `ALC pre-LD` +  `Age` + `Male sex` + `Bridging category`, data2, weights=weights, family = binomial ) %>% 
  tbl_regression(exponentiate=TRUE)

row1 <- tbl_merge(list(uv_tab, mv_tab), tab_spanner = c("**Univariate**", "**Multivariable**"))
row2 <- tbl_merge(list(uv_tab_IPTW, mv_tab_IPTW))
row3 <- tbl_merge(list(uv_tab_PSM, mv_tab_PSM))
tbl_stack(list(row1, row2, row3), group_header = c("UNWEIGHTED ANALYSIS", "IPTW ANALYSIS", "PSM ANALYSIS"), quiet=TRUE)

Characteristic	Univariate				Multivariable
	N	OR1	95% CI1	p-value	OR1	95% CI1	p-value
UNWEIGHTED ANALYSIS
Product	149
Liso-cel		—	—		—	—
Axi-cel		0.68	0.34, 1.35	0.3	0.61	0.26, 1.38	0.2
LDH pre-LD	148	0.14	0.02, 0.60	0.022	0.22	0.03, 0.89	0.083
ALC pre-LD	149	0.98	0.78, 1.19	0.8	1.02	0.73, 1.25	0.8
Largest lesion	149	0.86	0.77, 0.95	0.005	0.90	0.80, 1.00	0.064
Age	149	1.00	0.98, 1.03	>0.9	0.99	0.96, 1.02	0.5
Male sex	149	0.79	0.40, 1.53	0.5	0.73	0.34, 1.53	0.4
Bridging category	145
No bridging		—	—		—	—
CR/PR		0.92	0.35, 2.48	0.9	0.79	0.28, 2.25	0.6
SD/PD		0.35	0.15, 0.77	0.011	0.51	0.20, 1.30	0.2
IPTW ANALYSIS
Product	126
Liso-cel		—	—		—	—
Axi-cel		0.65	0.30, 1.39	0.3	0.52	0.21, 1.21	0.14
LDH pre-LD	125	0.09	0.01, 0.52	0.022	0.12	0.01, 0.72	0.052
ALC pre-LD	126	0.96	0.71, 1.21	0.7	1.02	0.66, 1.30	0.9
Largest lesion	126	0.92	0.82, 1.01	0.091	0.96	0.86, 1.08	0.5
Age	126	1.00	0.97, 1.03	>0.9	0.99	0.96, 1.02	0.6
Male sex	126	0.99	0.47, 2.07	>0.9	0.85	0.37, 1.93	0.7
Bridging category	123
No bridging		—	—		—	—
CR/PR		0.69	0.25, 1.87	0.5	0.68	0.24, 1.98	0.5
SD/PD		0.34	0.14, 0.82	0.017	0.42	0.15, 1.17	0.10
PSM ANALYSIS
Product	149
Liso-cel		—	—		—	—
Axi-cel		1.11	0.55, 2.20	0.8	1.02	0.45, 2.30	>0.9
LDH pre-LD	148	0.06	0.00, 0.44	0.016	0.07	0.00, 0.68	0.044
ALC pre-LD	149	0.96	0.75, 1.17	0.7	1.02	0.72, 1.27	0.9
Largest lesion	149	0.90	0.81, 1.00	0.060	0.98	0.86, 1.10	0.7
Age	149	1.03	1.00, 1.07	0.071	1.02	0.97, 1.06	0.5
Male sex	149	0.39	0.18, 0.78	0.009	0.37	0.16, 0.80	0.014
Bridging category	144
No bridging		—	—		—	—
CR/PR		0.91	0.36, 2.40	0.8	1.20	0.44, 3.52	0.7
SD/PD		0.34	0.14, 0.81	0.015	0.70	0.24, 2.01	0.5
1 OR = Odds Ratio, CI = Confidence Interval

Logistic CR forest plots

Show code

## Forest plot for MV
df_fp <- mv_tab$table_body %>%
  select(header_row, label, N_obs, N_event, estimate, conf.low, conf.high, p.value, ) %>%
  mutate(
    label = ifelse(header_row == TRUE | is.na(header_row),as.character(label), paste0("     ", as.character(label)) ),
    p.value = signif(p.value, 2),
    `OR (95% CI)` = ifelse(
      is.na(estimate),"",paste0(round(estimate, 2), " (", round(conf.low,2), " to ",round(conf.high,2), ") ")
    )
  ) %>%
  rename(Characteristic = label,
         N = N_obs,
         Events = N_event,
         `p value` = p.value)
forester(
  left_side_data = df_fp[2],
  right_side_data = df_fp[9:8],
  estimate = df_fp$estimate,
  ci_low = df_fp$conf.low,
  ci_high = df_fp$conf.high,
  estimate_col_name = "OR (95% CI)",
  xlim = c(.05, 3),
  null_line_at = 1,
  arrows = TRUE,
  x_scale_linear = FALSE,
  arrow_labels = c("Worse", "Better"),
  font_family = "sans",
  display = TRUE,
  nudge_x = .4,
  nudge_height = .06,
  stripe_colour = "#F5F5F5",
  file_path = "/Users/aportugu/Desktop/CD19 project/ASH abstract/forest/CR_mv_plot.png"
)

knitr::include_graphics("/Users/aportugu/Desktop/CD19 project/ASH abstract/forest/CR_mv_plot.png")

Show code

## Forest plot for MV IPTW
df_fp <- mv_tab_IPTW$table_body %>%
  select(header_row, label, N_obs, N_event, estimate, conf.low, conf.high, p.value, ) %>%
  mutate(
    label = ifelse(header_row == TRUE | is.na(header_row),as.character(label), paste0("     ", as.character(label)) ),
    p.value = signif(p.value, 2),
    `OR (95% CI)` = ifelse(
      is.na(estimate),"",paste0(round(estimate, 2), " (", round(conf.low,2), " to ",round(conf.high,2), ") ")
    )
  ) %>%
  rename(Characteristic = label,
         N = N_obs,
         Events = N_event,
         `p value` = p.value)
forester(
  left_side_data = df_fp[2],
  right_side_data = df_fp[9:8],
  estimate = df_fp$estimate,
  ci_low = df_fp$conf.low,
  ci_high = df_fp$conf.high,
  estimate_col_name = "OR (95% CI)",
  xlim = c(.05, 3),
  null_line_at = 1,
  arrows = TRUE,
  x_scale_linear = FALSE,
  arrow_labels = c("Worse", "Better"),
  font_family = "sans",
  display = TRUE,
  nudge_x = .4,
  nudge_height = .06,
  stripe_colour = "#F5F5F5",
  file_path = "/Users/aportugu/Desktop/CD19 project/ASH abstract/forest/CR_mv_plot_IPTW.png"
)

knitr::include_graphics("/Users/aportugu/Desktop/CD19 project/ASH abstract/forest/CR_mv_plot_IPTW.png")

Show code

## Forest plot for MV PSM
df_fp <- mv_tab_PSM$table_body %>%
  select(header_row, label, N_obs, N_event, estimate, conf.low, conf.high, p.value, ) %>%
  mutate(
    label = ifelse(header_row == TRUE | is.na(header_row),as.character(label), paste0("     ", as.character(label)) ),
    p.value = signif(p.value, 2),
    `OR (95% CI)` = ifelse(
      is.na(estimate),"",paste0(round(estimate, 2), " (", round(conf.low,2), " to ",round(conf.high,2), ") ")
    )
  ) %>%
  rename(Characteristic = label,
         N = N_obs,
         Events = N_event,
         `p value` = p.value)
forester(
  left_side_data = df_fp[2],
  right_side_data = df_fp[9:8],
  estimate = df_fp$estimate,
  ci_low = df_fp$conf.low,
  ci_high = df_fp$conf.high,
  estimate_col_name = "OR (95% CI)",
  xlim = c(.05, 3),
  null_line_at = 1,
  arrows = TRUE,
  x_scale_linear = FALSE,
  arrow_labels = c("Worse", "Better"),
  font_family = "sans",
  display = TRUE,
  nudge_x = .4,
  nudge_height = .06,
  stripe_colour = "#F5F5F5",
  file_path = "/Users/aportugu/Desktop/CD19 project/ASH abstract/forest/CR_mv_plot_PSM.png"
)

knitr::include_graphics("/Users/aportugu/Desktop/CD19 project/ASH abstract/forest/CR_mv_plot_PSM.png")

Toxicity

Logistic Model - ICANS grade 1+

Show code

theme_gtsummary_compact()
preds <- c("ICANS.any","Product", "LDH pre-LD","ALC pre-LD", "Largest lesion", "Age", "Male sex", "CMV", "Bridging category") 

data0 <- dataset %>%
  mutate(
    Product = factor(recode(CAR.T.Product.Type, "Yescarta" = "Axi-cel","Breyanzi" = "Liso-cel"), levels = c("Liso-cel","Axi-cel") ),
    "Largest lesion" = Largest.lesion,
    "HCT-CI" = HCT.CI,
    "ALC pre-LD" = ALC.preLD,
    "LDH pre-LD" = LDH.preLD/1000,
    "Male sex" = ifelse(Sex == "M",1,0),
    ICANS.any=ifelse(ICANS.grade == 0,0,1),
    "Bridging category" = factor(recode(Bridging.response.category, "0" = "No bridging", "1" = "CR/PR", "2" = "SD/PD"), levels = c("No bridging","CR/PR","SD/PD"))
  )

data1 <- IPTW.data %>%
  mutate(
    Product = factor(recode(CAR.T.Product.Type, "Yescarta" = "Axi-cel","Breyanzi" = "Liso-cel"), levels = c("Liso-cel","Axi-cel") ),
    "Largest lesion" = Largest.lesion,
    "HCT-CI" = HCT.CI,
    "ALC pre-LD" = ALC.preLD,
    "LDH pre-LD" = LDH.preLD/1000,
    "Male sex" = ifelse(Sex == "M",1,0),
    ICANS.any=ifelse(ICANS.grade == 0,0,1),
    "Bridging category" = factor(recode(Bridging.response.category, "0" = "No bridging", "1" = "CR/PR", "2" = "SD/PD"), levels = c("No bridging","CR/PR","SD/PD"))
  )

data2 <- m.data %>%
  mutate(
    Product = factor(recode(CAR.T.Product.Type, "Yescarta" = "Axi-cel","Breyanzi" = "Liso-cel"), levels = c("Liso-cel","Axi-cel") ),
    "Largest lesion" = Largest.lesion,
    "HCT-CI" = HCT.CI,
    "ALC pre-LD" = ALC.preLD,
    "LDH pre-LD" = LDH.preLD/1000,
    "Male sex" = ifelse(Sex == "M",1,0),
    ICANS.any=ifelse(ICANS.grade == 0,0,1),
    "Bridging category" = factor(recode(Bridging.response.category, "0" = "No bridging", "1" = "CR/PR", "2" = "SD/PD"), levels = c("No bridging","CR/PR","SD/PD"))
  )

uv_tab <- tbl_uvregression(
  data0[c(preds)],
  method = glm,
  y = ICANS.any,
  method.args = list(family = binomial),
  exponentiate = TRUE
)


mv_tab<-glm( ICANS.any ~ Product + `Largest lesion` + `ALC pre-LD` + Age + `Male sex` + `LDH pre-LD` + `Bridging category`, data0, family = binomial ) %>%
  tbl_regression(exponentiate=TRUE) 

uv_tab_IPTW <- tbl_uvregression(
  data1[c(preds)],
  method = glm,
  y = ICANS.any,
  method.args = list(family = binomial, weights = data1$weights),
  exponentiate = TRUE
)

mv_tab_IPTW<-glm( ICANS.any ~ Product + `Largest lesion` + `ALC pre-LD` + Age + `Male sex` + `LDH pre-LD` + `Bridging category`, data1, family = binomial, weights=weights ) %>%
  tbl_regression(exponentiate=TRUE) 


uv_tab_PSM <- tbl_uvregression(
  data2[c(preds)],
  method = glm,
  y = ICANS.any,
  method.args = list(family = binomial, weights = data2$weights),
  exponentiate = TRUE
)

mv_tab_PSM<-glm( ICANS.any ~ Product + `Largest lesion` + `ALC pre-LD` + Age + `Male sex` + `LDH pre-LD` + `Bridging category`, data2, family = binomial, weights=weights ) %>%
  tbl_regression(exponentiate=TRUE) 

row1 <- tbl_merge(list(uv_tab, mv_tab), tab_spanner = c("**Univariate**", "**Multivariable**"))
row2 <- tbl_merge(list(uv_tab_IPTW, mv_tab_IPTW))
row3 <- tbl_merge(list(uv_tab_PSM, mv_tab_PSM))
tbl_stack(list(row1, row2, row3), group_header = c("UNWEIGHTED ANALYSIS", "IPTW ANALYSIS", "PSM ANALYSIS"), quiet=TRUE)

Characteristic	Univariate				Multivariable
	N	OR1	95% CI1	p-value	OR1	95% CI1	p-value
UNWEIGHTED ANALYSIS
Product	149
Liso-cel		—	—		—	—
Axi-cel		2.84	1.39, 6.06	0.005	3.08	1.28, 7.84	0.014
LDH pre-LD	148	12.9	2.57, 124	0.009	8.15	1.63, 68.9	0.024
ALC pre-LD	149	1.29	0.96, 2.17	0.3	1.81	1.04, 3.36	0.043
Largest lesion	149	1.18	1.07, 1.32	0.002	1.18	1.05, 1.33	0.009
Age	149	1.01	0.98, 1.04	0.5	1.03	1.00, 1.06	0.072
Male sex	149	1.56	0.80, 3.10	0.2	1.57	0.72, 3.51	0.3
CMV	149	14.5	5.90, 41.4	<0.001
Bridging category	145
No bridging		—	—		—	—
CR/PR		0.63	0.22, 1.67	0.4	0.71	0.22, 2.17	0.6
SD/PD		1.57	0.72, 3.46	0.3	0.97	0.35, 2.62	>0.9
IPTW ANALYSIS
Product	126
Liso-cel		—	—		—	—
Axi-cel		1.75	0.82, 3.83	0.2	1.70	0.70, 4.28	0.2
LDH pre-LD	125	16.7	2.56, 239	0.014	11.2	1.73, 146	0.028
ALC pre-LD	126	1.41	0.96, 2.63	0.2	1.89	1.01, 3.88	0.059
Largest lesion	126	1.19	1.07, 1.35	0.002	1.20	1.06, 1.38	0.005
Age	126	1.01	0.98, 1.04	0.6	1.02	0.99, 1.06	0.3
Male sex	126	1.07	0.51, 2.25	0.9	1.15	0.49, 2.72	0.8
CMV	126	15.9	5.99, 51.0	<0.001
Bridging category	123
No bridging		—	—		—	—
CR/PR		0.60	0.20, 1.63	0.3	0.56	0.17, 1.72	0.3
SD/PD		1.47	0.63, 3.48	0.4	0.76	0.25, 2.23	0.6
PSM ANALYSIS
Product	149
Liso-cel		—	—		—	—
Axi-cel		1.64	0.80, 3.51	0.2	1.65	0.67, 4.31	0.3
LDH pre-LD	148	25.1	3.27, 368	0.007	9.80	1.12, 175	0.076
ALC pre-LD	149	1.29	0.97, 2.22	0.3	2.12	1.17, 4.13	0.017
Largest lesion	149	1.28	1.13, 1.47	<0.001	1.28	1.11, 1.50	0.002
Age	149	0.97	0.93, 1.00	0.061	0.97	0.93, 1.01	0.2
Male sex	149	1.36	0.68, 2.79	0.4	1.10	0.48, 2.54	0.8
CMV	149	26.0	9.63, 86.1	<0.001
Bridging category	144
No bridging		—	—		—	—
CR/PR		0.49	0.15, 1.34	0.2	0.43	0.10, 1.43	0.2
SD/PD		2.18	0.93, 5.19	0.073	0.78	0.23, 2.43	0.7
1 OR = Odds Ratio, CI = Confidence Interval

Logistic Model - ICANS grade 2+

Show code

theme_gtsummary_compact()
preds <- c("ICANS2.4","Product", "LDH pre-LD","ALC pre-LD", "Largest lesion", "Age", "Male sex", "Bridging category") 

data0 <- dataset %>%
  mutate(
    Product = factor(recode(CAR.T.Product.Type, "Yescarta" = "Axi-cel","Breyanzi" = "Liso-cel"), levels = c("Liso-cel","Axi-cel") ),
    "Largest lesion" = Largest.lesion,
    "HCT-CI" = HCT.CI,
    "ALC pre-LD" = ALC.preLD,
    "LDH pre-LD" = LDH.preLD/1000,
    "Male sex" = ifelse(Sex == "M",1,0),
    ICANS2.4=ifelse(ICANS.grade == 0 | ICANS.grade == 1 ,0,1),
    "Bridging category" = factor(recode(Bridging.response.category, "0" = "No bridging", "1" = "CR/PR", "2" = "SD/PD"), levels = c("No bridging","CR/PR","SD/PD"))
  )

data1 <- IPTW.data %>%
  mutate(
    Product = factor(recode(CAR.T.Product.Type, "Yescarta" = "Axi-cel","Breyanzi" = "Liso-cel"), levels = c("Liso-cel","Axi-cel") ),
    "Largest lesion" = Largest.lesion,
    "HCT-CI" = HCT.CI,
    "ALC pre-LD" = ALC.preLD,
    "LDH pre-LD" = LDH.preLD/1000,
    "Male sex" = ifelse(Sex == "M",1,0),
    ICANS2.4=ifelse(ICANS.grade == 0 | ICANS.grade == 1 ,0,1),
    "Bridging category" = factor(recode(Bridging.response.category, "0" = "No bridging", "1" = "CR/PR", "2" = "SD/PD"), levels = c("No bridging","CR/PR","SD/PD"))
  )

data2 <- m.data %>%
  mutate(
    Product = factor(recode(CAR.T.Product.Type, "Yescarta" = "Axi-cel","Breyanzi" = "Liso-cel"), levels = c("Liso-cel","Axi-cel") ),
    "Largest lesion" = Largest.lesion,
    "HCT-CI" = HCT.CI,
    "ALC pre-LD" = ALC.preLD,
    "LDH pre-LD" = LDH.preLD/1000,
    "Male sex" = ifelse(Sex == "M",1,0),
    ICANS2.4=ifelse(ICANS.grade == 0 | ICANS.grade == 1 ,0,1),
    "Bridging category" = factor(recode(Bridging.response.category, "0" = "No bridging", "1" = "CR/PR", "2" = "SD/PD"), levels = c("No bridging","CR/PR","SD/PD"))
  )

uv_tab <- tbl_uvregression(
  data0[c(preds)],
  method = glm,
  y = ICANS2.4,
  method.args = list(family = binomial),
  exponentiate = TRUE
)


mv_tab<-glm( ICANS2.4 ~ Product + `Largest lesion` + `ALC pre-LD` + Age + `Male sex` + `LDH pre-LD` + `Bridging category`, data0, family = binomial ) %>%
  tbl_regression(exponentiate=TRUE) 

uv_tab_IPTW <- tbl_uvregression(
  data1[c(preds)],
  method = glm,
  y = ICANS2.4,
  method.args = list(family = binomial, weights = data1$weights),
  exponentiate = TRUE
)

mv_tab_IPTW<-glm( ICANS2.4 ~ Product + `Largest lesion` + `ALC pre-LD` + Age + `Male sex` + `LDH pre-LD`+ `Bridging category`, data1, family = binomial, weights=weights ) %>%
  tbl_regression(exponentiate=TRUE) 


uv_tab_PSM <- tbl_uvregression(
  data2[c(preds)],
  method = glm,
  y = ICANS2.4,
  method.args = list(family = binomial, weights = data2$weights),
  exponentiate = TRUE
)

mv_tab_PSM<-glm( ICANS2.4 ~ Product + `Largest lesion` + `ALC pre-LD` + Age + `Male sex` + `LDH pre-LD`+ `Bridging category`, data2, family = binomial, weights=weights ) %>%
  tbl_regression(exponentiate=TRUE) 

row1 <- tbl_merge(list(uv_tab, mv_tab), tab_spanner = c("**Univariate**", "**Multivariable**"))
row2 <- tbl_merge(list(uv_tab_IPTW, mv_tab_IPTW))
row3 <- tbl_merge(list(uv_tab_PSM, mv_tab_PSM))
tbl_stack(list(row1, row2, row3), group_header = c("UNWEIGHTED ANALYSIS", "IPTW ANALYSIS", "PSM ANALYSIS"), quiet=TRUE)

Characteristic	Univariate				Multivariable
	N	OR1	95% CI1	p-value	OR1	95% CI1	p-value
UNWEIGHTED ANALYSIS
Product	149
Liso-cel		—	—		—	—
Axi-cel		2.15	1.02, 4.75	0.050	1.88	0.78, 4.80	0.2
LDH pre-LD	148	6.62	1.82, 34.8	0.010	6.13	1.50, 35.0	0.021
ALC pre-LD	149	1.20	0.96, 1.92	0.3	1.49	0.95, 2.63	0.2
Largest lesion	149	1.15	1.04, 1.27	0.007	1.16	1.03, 1.31	0.019
Age	149	1.01	0.98, 1.04	0.5	1.03	0.99, 1.06	0.12
Male sex	149	1.81	0.89, 3.80	0.11	2.03	0.90, 4.78	0.095
Bridging category	145
No bridging		—	—		—	—
CR/PR		0.55	0.17, 1.55	0.3	0.60	0.16, 1.90	0.4
SD/PD		1.25	0.56, 2.75	0.6	0.63	0.21, 1.70	0.4
IPTW ANALYSIS
Product	126
Liso-cel		—	—		—	—
Axi-cel		1.29	0.59, 2.91	0.5	1.09	0.44, 2.75	0.9
LDH pre-LD	125	8.14	1.81, 59.0	0.015	7.60	1.52, 58.2	0.026
ALC pre-LD	126	1.22	0.93, 2.15	0.4	1.55	0.91, 3.01	0.2
Largest lesion	126	1.17	1.05, 1.31	0.005	1.19	1.05, 1.36	0.009
Age	126	1.01	0.98, 1.04	0.7	1.01	0.98, 1.05	0.4
Male sex	126	1.31	0.61, 2.89	0.5	1.53	0.64, 3.83	0.3
Bridging category	123
No bridging		—	—		—	—
CR/PR		0.57	0.17, 1.65	0.3	0.54	0.14, 1.71	0.3
SD/PD		1.25	0.52, 2.94	0.6	0.53	0.16, 1.57	0.3
PSM ANALYSIS
Product	149
Liso-cel		—	—		—	—
Axi-cel		1.67	0.79, 3.71	0.2	1.64	0.65, 4.46	0.3
LDH pre-LD	148	13.5	2.26, 137	0.012	6.27	0.89, 82.3	0.11
ALC pre-LD	149	1.24	0.97, 2.07	0.3	1.98	1.08, 3.75	0.028
Largest lesion	149	1.25	1.11, 1.43	<0.001	1.25	1.09, 1.46	0.003
Age	149	0.96	0.92, 1.00	0.032	0.96	0.92, 1.00	0.069
Male sex	149	1.47	0.71, 3.11	0.3	1.24	0.53, 2.96	0.6
Bridging category	144
No bridging		—	—		—	—
CR/PR		0.39	0.10, 1.17	0.12	0.33	0.07, 1.21	0.13
SD/PD		1.95	0.82, 4.62	0.13	0.64	0.19, 1.99	0.5
1 OR = Odds Ratio, CI = Confidence Interval

Logistic Model - ICANS grade 3+

Show code

theme_gtsummary_compact()
preds <- c("ICANS3.4","Product", "LDH pre-LD","ALC pre-LD", "Largest lesion", "Age", "Male sex", "CMV", "Bridging category") 

data0 <- dataset %>%
  mutate(
    Product = factor(recode(CAR.T.Product.Type, "Yescarta" = "Axi-cel","Breyanzi" = "Liso-cel"), levels = c("Liso-cel","Axi-cel") ),
    "Largest lesion" = Largest.lesion,
    "HCT-CI" = HCT.CI,
    "ALC pre-LD" = ALC.preLD,
    "LDH pre-LD" = LDH.preLD/1000,
    "Male sex" = ifelse(Sex == "M",1,0),
    ICANS3.4=ifelse(ICANS.grade == 3 | ICANS.grade == 4,1,0),
    "Bridging category" = factor(recode(Bridging.response.category, "0" = "No bridging", "1" = "CR/PR", "2" = "SD/PD"), levels = c("No bridging","CR/PR","SD/PD"))
  )

data1 <- IPTW.data %>%
  mutate(
    Product = factor(recode(CAR.T.Product.Type, "Yescarta" = "Axi-cel","Breyanzi" = "Liso-cel"), levels = c("Liso-cel","Axi-cel") ),
    "Largest lesion" = Largest.lesion,
    "HCT-CI" = HCT.CI,
    "ALC pre-LD" = ALC.preLD,
    "LDH pre-LD" = LDH.preLD/1000,
    "Male sex" = ifelse(Sex == "M",1,0),
    ICANS3.4=ifelse(ICANS.grade == 3 | ICANS.grade == 4,1,0),
    "Bridging category" = factor(recode(Bridging.response.category, "0" = "No bridging", "1" = "CR/PR", "2" = "SD/PD"), levels = c("No bridging","CR/PR","SD/PD"))
  )

data2 <- m.data %>%
  mutate(
    Product = factor(recode(CAR.T.Product.Type, "Yescarta" = "Axi-cel","Breyanzi" = "Liso-cel"), levels = c("Liso-cel","Axi-cel") ),
    "Largest lesion" = Largest.lesion,
    "HCT-CI" = HCT.CI,
    "ALC pre-LD" = ALC.preLD,
    "LDH pre-LD" = LDH.preLD/1000,
    "Male sex" = ifelse(Sex == "M",1,0),
    ICANS3.4=ifelse(ICANS.grade == 3 | ICANS.grade == 4,1,0),
    "Bridging category" = factor(recode(Bridging.response.category, "0" = "No bridging", "1" = "CR/PR", "2" = "SD/PD"), levels = c("No bridging","CR/PR","SD/PD"))
  )

uv_tab <- tbl_uvregression(
  data0[c(preds)],
  method = glm,
  y = ICANS3.4,
  method.args = list(family = binomial),
  exponentiate = TRUE
)

mv_tab<-glm( ICANS3.4 ~ Product + `Largest lesion` + `ALC pre-LD` + Age + `Male sex` + `LDH pre-LD` + `Bridging category`, data0, family = binomial ) %>% tbl_regression(exponentiate=TRUE) 

uv_tab_IPTW <- tbl_uvregression(
  data1[c(preds)],
  method = glm,
  y = ICANS3.4,
  method.args = list(family = binomial, weights = data1$weights),
  exponentiate = TRUE
)

mv_tab_IPTW<-glm( ICANS3.4 ~ Product + `Largest lesion` + `ALC pre-LD` + Age + `Male sex` + `LDH pre-LD` + `Bridging category`, data1, family = binomial, weights=weights ) %>%
  tbl_regression(exponentiate=TRUE) 

uv_tab_PSM <- tbl_uvregression(
  data2[c(preds)],
  method = glm,
  y = ICANS3.4,
  method.args = list(family = binomial, weights = data2$weights),
  exponentiate = TRUE
)

mv_tab_PSM<-glm( ICANS3.4 ~ Product + `Largest lesion` + `ALC pre-LD` + Age + `Male sex` + `LDH pre-LD` + `Bridging category`, data2, family = binomial, weights=weights ) %>%
  tbl_regression(exponentiate=TRUE) 

row1 <- tbl_merge(list(uv_tab, mv_tab), tab_spanner = c("**Univariate**", "**Multivariable**"))
row2 <- tbl_merge(list(uv_tab_IPTW, mv_tab_IPTW))
row3 <- tbl_merge(list(uv_tab_PSM, mv_tab_PSM))
tbl_stack(list(row1, row2, row3), group_header = c("UNWEIGHTED ANALYSIS", "IPTW ANALYSIS", "PSM ANALYSIS"), quiet=TRUE)

Characteristic	Univariate				Multivariable
	N	OR1	95% CI1	p-value	OR1	95% CI1	p-value
UNWEIGHTED ANALYSIS
Product	149
Liso-cel		—	—		—	—
Axi-cel		1.01	0.43, 2.54	>0.9	0.85	0.30, 2.45	0.8
LDH pre-LD	148	2.59	1.09, 8.63	0.12	2.58	1.08, 10.6	0.2
ALC pre-LD	149	0.97	0.59, 1.19	0.8	0.94	0.63, 1.15	0.6
Largest lesion	149	1.15	1.03, 1.29	0.012	1.15	1.01, 1.31	0.039
Age	149	1.02	0.98, 1.05	0.3	1.03	0.99, 1.08	0.2
Male sex	149	1.92	0.78, 5.20	0.2	2.82	1.01, 9.18	0.061
CMV	149	17.8	6.75, 54.0	<0.001
Bridging category	145
No bridging		—	—		—	—
CR/PR		1.36	0.35, 4.41	0.6	1.56	0.37, 5.69	0.5
SD/PD		2.22	0.85, 5.67	0.10	1.33	0.40, 4.12	0.6
IPTW ANALYSIS
Product	126
Liso-cel		—	—		—	—
Axi-cel		0.60	0.24, 1.53	0.3	0.48	0.17, 1.37	0.2
LDH pre-LD	125	3.34	1.06, 13.6	0.075	3.65	1.09, 18.2	0.094
ALC pre-LD	126	0.97	0.55, 1.22	0.8	0.92	0.59, 1.16	0.5
Largest lesion	126	1.17	1.04, 1.32	0.008	1.15	1.01, 1.33	0.037
Age	126	1.00	0.97, 1.04	0.8	1.02	0.97, 1.06	0.5
Male sex	126	1.41	0.55, 3.91	0.5	2.08	0.71, 6.99	0.2
CMV	126	17.3	6.21, 56.2	<0.001
Bridging category	123
No bridging		—	—		—	—
CR/PR		1.28	0.33, 4.24	0.7	1.24	0.29, 4.50	0.8
SD/PD		2.01	0.72, 5.47	0.2	0.94	0.24, 3.20	>0.9
PSM ANALYSIS
Product	149
Liso-cel		—	—		—	—
Axi-cel		0.95	0.40, 2.41	>0.9	0.83	0.30, 2.38	0.7
LDH pre-LD	148	5.02	1.04, 28.6	0.051	2.95	0.62, 22.9	0.3
ALC pre-LD	149	0.99	0.65, 1.21	>0.9	0.94	0.68, 1.16	0.6
Largest lesion	149	1.21	1.07, 1.38	0.003	1.17	1.02, 1.35	0.028
Age	149	0.97	0.93, 1.01	0.13	0.98	0.93, 1.03	0.4
Male sex	149	1.70	0.70, 4.55	0.3	1.54	0.57, 4.50	0.4
CMV	149	12.6	4.95, 34.4	<0.001
Bridging category	144
No bridging		—	—		—	—
CR/PR		0.89	0.21, 2.97	0.9	0.81	0.18, 2.92	0.8
SD/PD		2.50	0.92, 6.60	0.067	1.03	0.28, 3.44	>0.9
1 OR = Odds Ratio, CI = Confidence Interval

Logistic Model - CRS grade 1+

Show code

theme_gtsummary_compact()
preds <- c("CRS.any","Product", "LDH pre-LD","ALC pre-LD", "Largest lesion", "Age", "Male sex", "CMV", "Bridging category") 

data0 <- dataset %>%
  mutate(
    Product = factor(recode(CAR.T.Product.Type, "Yescarta" = "Axi-cel","Breyanzi" = "Liso-cel"), levels = c("Liso-cel","Axi-cel") ),
    "Largest lesion" = Largest.lesion,
    "HCT-CI" = HCT.CI,
    "ALC pre-LD" = ALC.preLD,
    "LDH pre-LD" = LDH.preLD/1000,
    "Male sex" = ifelse(Sex == "M",1,0),
    CRS.any=ifelse(CRS.grade == 0,0,1),
    "Bridging category" = factor(recode(Bridging.response.category, "0" = "No bridging", "1" = "CR/PR", "2" = "SD/PD"), levels = c("No bridging","CR/PR","SD/PD"))
  )

data1 <- IPTW.data %>%
  mutate(
    Product = factor(recode(CAR.T.Product.Type, "Yescarta" = "Axi-cel","Breyanzi" = "Liso-cel"), levels = c("Liso-cel","Axi-cel") ),
    "Largest lesion" = Largest.lesion,
    "HCT-CI" = HCT.CI,
    "ALC pre-LD" = ALC.preLD,
    "LDH pre-LD" = LDH.preLD/1000,
    "Male sex" = ifelse(Sex == "M",1,0),
    CRS.any=ifelse(CRS.grade == 0,0,1),
    "Bridging category" = factor(recode(Bridging.response.category, "0" = "No bridging", "1" = "CR/PR", "2" = "SD/PD"), levels = c("No bridging","CR/PR","SD/PD"))
  )

data2 <- m.data %>%
  mutate(
    Product = factor(recode(CAR.T.Product.Type, "Yescarta" = "Axi-cel","Breyanzi" = "Liso-cel"), levels = c("Liso-cel","Axi-cel") ),
    "Largest lesion" = Largest.lesion,
    "HCT-CI" = HCT.CI,
    "ALC pre-LD" = ALC.preLD,
    "LDH pre-LD" = LDH.preLD/1000,
    "Male sex" = ifelse(Sex == "M",1,0),
    CRS.any=ifelse(CRS.grade == 0,0,1),
    "Bridging category" = factor(recode(Bridging.response.category, "0" = "No bridging", "1" = "CR/PR", "2" = "SD/PD"), levels = c("No bridging","CR/PR","SD/PD"))
  )

uv_tab <- tbl_uvregression(
  data0[c(preds)],
  method = glm,
  y = CRS.any,
  method.args = list(family = binomial),
  exponentiate = TRUE
)

mv_tab<-glm( CRS.any ~ Product + `Largest lesion` + `ALC pre-LD` + Age + `Male sex` + `LDH pre-LD` + `Bridging category`, data0, family = binomial ) %>% tbl_regression(exponentiate=TRUE) 

uv_tab_IPTW <- tbl_uvregression(
  data1[c(preds)],
  method = glm,
  y = CRS.any,
  method.args = list(family = binomial, weights = data1$weights),
  exponentiate = TRUE
)

mv_tab_IPTW<-glm( CRS.any ~ Product + `Largest lesion` + `ALC pre-LD` + Age + `Male sex` + `LDH pre-LD` + `Bridging category`, data1, family = binomial, weights=weights ) %>%
  tbl_regression(exponentiate=TRUE) 

uv_tab_PSM <- tbl_uvregression(
  data2[c(preds)],
  method = glm,
  y = CRS.any,
  method.args = list(family = binomial, weights = data2$weights),
  exponentiate = TRUE
)

mv_tab_PSM<-glm( CRS.any ~ Product + `Largest lesion` + `ALC pre-LD` + Age + `Male sex` + `LDH pre-LD` + `Bridging category`, data2, family = binomial, weights=weights ) %>%
  tbl_regression(exponentiate=TRUE) 

row1 <- tbl_merge(list(uv_tab, mv_tab), tab_spanner = c("**Univariate**", "**Multivariable**"))
row2 <- tbl_merge(list(uv_tab_IPTW, mv_tab_IPTW))
row3 <- tbl_merge(list(uv_tab_PSM, mv_tab_PSM))
tbl_stack(list(row1, row2, row3), group_header = c("UNWEIGHTED ANALYSIS", "IPTW ANALYSIS", "PSM ANALYSIS"), quiet=TRUE)

Characteristic	Univariate				Multivariable
	N	OR1	95% CI1	p-value	OR1	95% CI1	p-value
UNWEIGHTED ANALYSIS
Product	149
Liso-cel		—	—		—	—
Axi-cel		4.50	1.95, 10.8	<0.001	4.71	1.60, 14.9	0.006
LDH pre-LD	148	89.4	2.31, 25,644	0.065	89.0	1.37, 47,171	0.10
ALC pre-LD	149	1.69	0.94, 4.14	0.2	3.35	1.37, 11.8	0.028
Largest lesion	149	1.28	1.09, 1.56	0.006	1.30	1.09, 1.61	0.007
Age	149	0.98	0.94, 1.01	0.2	0.99	0.94, 1.03	0.5
Male sex	149	0.85	0.35, 1.95	0.7	0.52	0.17, 1.46	0.2
CMV	149	14.0	2.82, 254	0.011
Bridging category	145
No bridging		—	—		—	—
CR/PR		0.28	0.10, 0.83	0.019	0.23	0.06, 0.85	0.027
SD/PD		0.72	0.27, 2.07	0.5	0.30	0.08, 1.08	0.065
IPTW ANALYSIS
Product	126
Liso-cel		—	—		—	—
Axi-cel		2.69	1.12, 6.54	0.027	3.89	1.26, 13.1	0.021
LDH pre-LD	125	667	5.16, 671,492	0.033	433	2.95, 533,357	0.051
ALC pre-LD	126	2.05	0.96, 5.88	0.13	3.30	1.20, 14.0	0.057
Largest lesion	126	1.26	1.07, 1.53	0.012	1.30	1.09, 1.62	0.007
Age	126	0.97	0.93, 1.01	0.2	0.98	0.93, 1.02	0.3
Male sex	126	0.53	0.18, 1.34	0.2	0.33	0.08, 1.07	0.080
CMV	126	22.5	3.52, 1,160	0.014
Bridging category	123
No bridging		—	—		—	—
CR/PR		0.22	0.07, 0.67	0.007	0.16	0.04, 0.63	0.010
SD/PD		0.68	0.23, 2.14	0.5	0.24	0.06, 1.00	0.052
PSM ANALYSIS
Product	149
Liso-cel		—	—		—	—
Axi-cel		2.11	0.99, 4.49	0.053	3.53	1.27, 10.5	0.018
LDH pre-LD	148	126	3.12, 21,564	0.034	522	5.46, 178,178	0.018
ALC pre-LD	149	3.62	1.54, 9.78	0.006	4.78	1.78, 18.6	0.010
Largest lesion	149	1.21	1.05, 1.45	0.018	1.29	1.08, 1.59	0.009
Age	149	1.00	0.96, 1.04	>0.9	0.99	0.94, 1.04	0.7
Male sex	149	0.34	0.14, 0.77	0.013	0.22	0.07, 0.65	0.009
CMV	149	16.7	3.40, 303	0.006
Bridging category	144
No bridging		—	—		—	—
CR/PR		0.17	0.06, 0.44	<0.001	0.16	0.04, 0.52	0.003
SD/PD		0.81	0.30, 2.39	0.7	0.51	0.13, 1.98	0.3
1 OR = Odds Ratio, CI = Confidence Interval

Logistic Model - CRS grade 2+

Show code

theme_gtsummary_compact()
preds <- c("CRS2.4","Product", "LDH pre-LD","ALC pre-LD", "Largest lesion", "Age", "Male sex", "CMV", "Bridging category") 

data0 <- dataset %>%
  mutate(
    Product = factor(recode(CAR.T.Product.Type, "Yescarta" = "Axi-cel","Breyanzi" = "Liso-cel"), levels = c("Liso-cel","Axi-cel") ),
    "Largest lesion" = Largest.lesion,
    "HCT-CI" = HCT.CI,
    "ALC pre-LD" = ALC.preLD,
    "LDH pre-LD" = LDH.preLD/1000,
    "Male sex" = ifelse(Sex == "M",1,0),
    CRS2.4=ifelse(CRS.grade == 0 | CRS.grade == 1 ,0,1),
    "Bridging category" = factor(recode(Bridging.response.category, "0" = "No bridging", "1" = "CR/PR", "2" = "SD/PD"), levels = c("No bridging","CR/PR","SD/PD"))
  )

data1 <- IPTW.data %>%
  mutate(
    Product = factor(recode(CAR.T.Product.Type, "Yescarta" = "Axi-cel","Breyanzi" = "Liso-cel"), levels = c("Liso-cel","Axi-cel") ),
    "Largest lesion" = Largest.lesion,
    "HCT-CI" = HCT.CI,
    "ALC pre-LD" = ALC.preLD,
    "LDH pre-LD" = LDH.preLD/1000,
    "Male sex" = ifelse(Sex == "M",1,0),
    CRS2.4=ifelse(CRS.grade == 0 | CRS.grade == 1 ,0,1),
    "Bridging category" = factor(recode(Bridging.response.category, "0" = "No bridging", "1" = "CR/PR", "2" = "SD/PD"), levels = c("No bridging","CR/PR","SD/PD"))
  )

data2 <- m.data %>%
  mutate(
    Product = factor(recode(CAR.T.Product.Type, "Yescarta" = "Axi-cel","Breyanzi" = "Liso-cel"), levels = c("Liso-cel","Axi-cel") ),
    "Largest lesion" = Largest.lesion,
    "HCT-CI" = HCT.CI,
    "ALC pre-LD" = ALC.preLD,
    "LDH pre-LD" = LDH.preLD/1000,
    "Male sex" = ifelse(Sex == "M",1,0),
    CRS2.4=ifelse(CRS.grade == 0 | CRS.grade == 1 ,0,1),
    "Bridging category" = factor(recode(Bridging.response.category, "0" = "No bridging", "1" = "CR/PR", "2" = "SD/PD"), levels = c("No bridging","CR/PR","SD/PD"))
  )

uv_tab <- tbl_uvregression(
  data0[c(preds)],
  method = glm,
  y = CRS2.4,
  method.args = list(family = binomial),
  exponentiate = TRUE
)

mv_tab<-glm( CRS2.4 ~ Product + `Largest lesion` + `ALC pre-LD` + Age + `Male sex` + `LDH pre-LD` + `Bridging category`, data0, family = binomial ) %>% tbl_regression(exponentiate=TRUE) 

uv_tab_IPTW <- tbl_uvregression(
  data1[c(preds)],
  method = glm,
  y = CRS2.4,
  method.args = list(family = binomial, weights = data1$weights),
  exponentiate = TRUE
)

mv_tab_IPTW<-glm( CRS2.4 ~ Product + `Largest lesion` + `ALC pre-LD` + Age + `Male sex` + `LDH pre-LD` + `Bridging category`, data1, family = binomial, weights=weights ) %>%
  tbl_regression(exponentiate=TRUE) 

uv_tab_PSM <- tbl_uvregression(
  data2[c(preds)],
  method = glm,
  y = CRS2.4,
  method.args = list(family = binomial, weights = data2$weights),
  exponentiate = TRUE
)

mv_tab_PSM<-glm( CRS2.4 ~ Product + `Largest lesion` + `ALC pre-LD` + Age + `Male sex` + `LDH pre-LD` + `Bridging category`, data2, family = binomial, weights=weights ) %>%
  tbl_regression(exponentiate=TRUE) 

row1 <- tbl_merge(list(uv_tab, mv_tab), tab_spanner = c("**Univariate**", "**Multivariable**"))
row2 <- tbl_merge(list(uv_tab_IPTW, mv_tab_IPTW))
row3 <- tbl_merge(list(uv_tab_PSM, mv_tab_PSM))
tbl_stack(list(row1, row2, row3), group_header = c("UNWEIGHTED ANALYSIS", "IPTW ANALYSIS", "PSM ANALYSIS"), quiet=TRUE)

Characteristic	Univariate				Multivariable
	N	OR1	95% CI1	p-value	OR1	95% CI1	p-value
UNWEIGHTED ANALYSIS
Product	149
Liso-cel		—	—		—	—
Axi-cel		4.08	1.90, 9.46	<0.001	4.25	1.79, 11.0	0.002
LDH pre-LD	148	0.96	0.59, 1.29	0.8	0.88	0.47, 1.20	0.5
ALC pre-LD	149	0.94	0.66, 1.14	0.6	0.98	0.72, 1.19	0.9
Largest lesion	149	1.08	0.98, 1.19	0.11	1.08	0.97, 1.20	0.2
Age	149	0.99	0.96, 1.01	0.3	1.00	0.97, 1.03	>0.9
Male sex	149	0.84	0.43, 1.66	0.6	0.69	0.32, 1.45	0.3
CMV	149	2.13	1.03, 4.46	0.042
Bridging category	145
No bridging		—	—		—	—
CR/PR		1.14	0.42, 2.97	0.8	1.68	0.57, 4.98	0.3
SD/PD		1.21	0.55, 2.65	0.6	1.18	0.47, 2.96	0.7
IPTW ANALYSIS
Product	126
Liso-cel		—	—		—	—
Axi-cel		3.07	1.36, 7.45	0.009	3.19	1.32, 8.36	0.013
LDH pre-LD	125	1.00	0.58, 1.53	>0.9	0.91	0.47, 1.38	0.6
ALC pre-LD	126	0.97	0.67, 1.21	0.8	0.98	0.69, 1.24	0.9
Largest lesion	126	1.06	0.96, 1.17	0.3	1.07	0.95, 1.20	0.2
Age	126	0.99	0.96, 1.02	0.5	1.00	0.96, 1.03	0.8
Male sex	126	0.58	0.27, 1.21	0.15	0.49	0.21, 1.09	0.083
CMV	126	2.36	1.08, 5.22	0.032
Bridging category	123
No bridging		—	—		—	—
CR/PR		0.99	0.35, 2.69	>0.9	1.09	0.37, 3.17	0.9
SD/PD		1.16	0.49, 2.73	0.7	1.04	0.38, 2.84	>0.9
PSM ANALYSIS
Product	149
Liso-cel		—	—		—	—
Axi-cel		3.86	1.79, 8.95	<0.001	5.14	2.05, 14.3	<0.001
LDH pre-LD	148	1.20	0.40, 5.14	0.7	0.98	0.27, 6.34	>0.9
ALC pre-LD	149	1.00	0.78, 1.21	>0.9	0.97	0.73, 1.19	0.8
Largest lesion	149	1.00	0.90, 1.11	>0.9	1.02	0.89, 1.17	0.7
Age	149	1.00	0.96, 1.03	>0.9	0.99	0.94, 1.03	0.6
Male sex	149	0.27	0.13, 0.54	<0.001	0.17	0.07, 0.39	<0.001
CMV	149	2.11	0.98, 4.59	0.058
Bridging category	144
No bridging		—	—		—	—
CR/PR		0.25	0.07, 0.72	0.017	0.39	0.11, 1.22	0.13
SD/PD		0.90	0.38, 2.09	0.8	1.38	0.45, 4.23	0.6
1 OR = Odds Ratio, CI = Confidence Interval

Logistic Model - CRS grade 3+

Show code

theme_gtsummary_compact()
preds <- c("CRS3.4","Product", "LDH pre-LD","ALC pre-LD", "Largest lesion", "Age", "Male sex", "CMV", "Bridging category") 

data0 <- dataset %>%
  mutate(
    Product = factor(recode(CAR.T.Product.Type, "Yescarta" = "Axi-cel","Breyanzi" = "Liso-cel"), levels = c("Liso-cel","Axi-cel") ),
    "Largest lesion" = Largest.lesion,
    "HCT-CI" = HCT.CI,
    "ALC pre-LD" = ALC.preLD,
    "LDH pre-LD" = LDH.preLD/1000,
    "Male sex" = ifelse(Sex == "M",1,0),
    CRS3.4=ifelse(CRS.grade == 3 | CRS.grade == 4,1,0),
    "Bridging category" = factor(recode(Bridging.response.category, "0" = "No bridging", "1" = "CR/PR", "2" = "SD/PD"), levels = c("No bridging","CR/PR","SD/PD"))
  )

data1 <- IPTW.data %>%
  mutate(
    Product = factor(recode(CAR.T.Product.Type, "Yescarta" = "Axi-cel","Breyanzi" = "Liso-cel"), levels = c("Liso-cel","Axi-cel") ),
    "Largest lesion" = Largest.lesion,
    "HCT-CI" = HCT.CI,
    "ALC pre-LD" = ALC.preLD,
    "LDH pre-LD" = LDH.preLD/1000,
    "Male sex" = ifelse(Sex == "M",1,0),
    CRS3.4=ifelse(CRS.grade == 3 | CRS.grade == 4,1,0),
    "Bridging category" = factor(recode(Bridging.response.category, "0" = "No bridging", "1" = "CR/PR", "2" = "SD/PD"), levels = c("No bridging","CR/PR","SD/PD"))
  )

data2 <- m.data %>%
  mutate(
    Product = factor(recode(CAR.T.Product.Type, "Yescarta" = "Axi-cel","Breyanzi" = "Liso-cel"), levels = c("Liso-cel","Axi-cel") ),
    "Largest lesion" = Largest.lesion,
    "HCT-CI" = HCT.CI,
    "ALC pre-LD" = ALC.preLD,
    "LDH pre-LD" = LDH.preLD/1000,
    "Male sex" = ifelse(Sex == "M",1,0),
    CRS3.4=ifelse(CRS.grade == 3 | CRS.grade == 4,1,0),
    "Bridging category" = factor(recode(Bridging.response.category, "0" = "No bridging", "1" = "CR/PR", "2" = "SD/PD"), levels = c("No bridging","CR/PR","SD/PD"))
  )

uv_tab <- tbl_uvregression(
  data0[c(preds)],
  method = glm,
  y = CRS3.4,
  method.args = list(family = binomial),
  exponentiate = TRUE
)

mv_tab<-glm( CRS3.4 ~ Product + `Largest lesion` + `ALC pre-LD` + Age + `Male sex` + `LDH pre-LD` + `Bridging category`, data0, family = binomial ) %>% tbl_regression(exponentiate=TRUE) 

uv_tab_IPTW <- tbl_uvregression(
  data1[c(preds)],
  method = glm,
  y = CRS3.4,
  method.args = list(family = binomial, weights = data1$weights),
  exponentiate = TRUE
)

mv_tab_IPTW<-glm( CRS3.4 ~ Product + `Largest lesion` + `ALC pre-LD` + Age + `Male sex` + `LDH pre-LD` + `Bridging category`, data1, family = binomial, weights=weights ) %>%
  tbl_regression(exponentiate=TRUE) 

uv_tab_PSM <- tbl_uvregression(
  data2[c(preds)],
  method = glm,
  y = CRS3.4,
  method.args = list(family = binomial, weights = data2$weights),
  exponentiate = TRUE
)

mv_tab_PSM<-glm( CRS3.4 ~ Product + `Largest lesion` + `ALC pre-LD` + Age + `Male sex` + `LDH pre-LD` + `Bridging category`, data2, family = binomial, weights=weights ) %>%
  tbl_regression(exponentiate=TRUE) 

row1 <- tbl_merge(list(uv_tab, mv_tab), tab_spanner = c("**Univariate**", "**Multivariable**"))
row2 <- tbl_merge(list(uv_tab_IPTW, mv_tab_IPTW))
row3 <- tbl_merge(list(uv_tab_PSM, mv_tab_PSM))
tbl_stack(list(row1, row2, row3), group_header = c("UNWEIGHTED ANALYSIS", "IPTW ANALYSIS", "PSM ANALYSIS"), quiet=TRUE)

Characteristic	Univariate				Multivariable
	N	OR1	95% CI1	p-value	OR1	95% CI1	p-value
UNWEIGHTED ANALYSIS
Product	149
Liso-cel		—	—		—	—
Axi-cel		4.31	0.76, 81.1	0.2	2.87	0.45, 56.2	0.3
LDH pre-LD	148	1.10	0.53, 1.50	0.6	1.10	0.52, 1.57	0.6
ALC pre-LD	149	0.66	0.13, 1.20	0.5	0.62	0.11, 1.31	0.5
Largest lesion	149	1.02	0.83, 1.21	0.8	0.99	0.78, 1.22	>0.9
Age	149	0.98	0.94, 1.04	0.5	1.00	0.95, 1.06	>0.9
Male sex	149	2.20	0.51, 15.1	0.3	2.09	0.45, 16.5	0.4
CMV	149	2.23	0.53, 8.86	0.3
Bridging category	145
No bridging		—	—		—	—
CR/PR		0.00		>0.9	0.00		>0.9
SD/PD		0.68	0.10, 2.99	0.6	0.68	0.08, 3.61	0.7
IPTW ANALYSIS
Product	126
Liso-cel		—	—		—	—
Axi-cel		2.70	0.50, 35.6	0.3	2.07	0.36, 28.1	0.5
LDH pre-LD	125	1.10		0.7	1.15	0.42, 1.80	0.6
ALC pre-LD	126	0.86	0.18, 1.30	0.8	0.71	0.12, 1.39	0.7
Largest lesion	126	0.98	0.77, 1.18	0.9	0.97	0.73, 1.22	0.8
Age	126	0.99	0.94, 1.06	0.8	1.00	0.94, 1.07	>0.9
Male sex	126	2.38	0.47, 23.8	0.4	2.31	0.42, 27.0	0.4
CMV	126	1.99	0.42, 8.67	0.4
Bridging category	123
No bridging		—	—		—	—
CR/PR		0.00		>0.9	0.00		>0.9
SD/PD		0.50	0.05, 2.57	0.5	0.55	0.04, 3.55	0.6
PSM ANALYSIS
Product	149
Liso-cel		—	—		—	—
Axi-cel		2.21	0.32, 43.6	0.5	1.72	0.22, 37.4	0.6
LDH pre-LD	148	1.17		0.8	1.28		0.6
ALC pre-LD	149	1.02		>0.9	1.10		0.7
Largest lesion	149	0.91	0.60, 1.19	0.6	0.89	0.53, 1.26	0.6
Age	149	0.99	0.91, 1.10	0.9	0.98	0.90, 1.10	0.8
Male sex	149	0.95	0.16, 7.09	>0.9	1.24	0.19, 10.6	0.8
CMV	149	1.76	0.21, 10.5	0.5
Bridging category	144
No bridging		—	—		—	—
CR/PR		0.00		>0.9	0.00		>0.9
SD/PD		0.27	0.00, 2.87	0.4	0.27	0.00, 4.48	0.5
1 OR = Odds Ratio, CI = Confidence Interval

Medication administration

Tocilizumab

Show code

theme_gtsummary_compact()
preds <- c("toci.doses","Product", "LDH pre-LD","ALC pre-LD", "Largest lesion", "Age", "Male sex", "CMV", "Bridging category") 

data0 <- dataset %>%
  mutate(
    Product = factor(recode(CAR.T.Product.Type, "Yescarta" = "Axi-cel","Breyanzi" = "Liso-cel"), levels = c("Liso-cel","Axi-cel") ),
    "Largest lesion" = Largest.lesion,
    "HCT-CI" = HCT.CI,
    "ALC pre-LD" = ALC.preLD,
    "LDH pre-LD" = LDH.preLD/1000,
    "Male sex" = ifelse(Sex == "M",1,0),
    CRS2.4=ifelse(CRS.grade == 0 | CRS.grade == 1 ,0,1),
    "Bridging category" = factor(recode(Bridging.response.category, "0" = "No bridging", "1" = "CR/PR", "2" = "SD/PD"), levels = c("No bridging","CR/PR","SD/PD"))
  )

data1 <- IPTW.data %>%
  mutate(
    Product = factor(recode(CAR.T.Product.Type, "Yescarta" = "Axi-cel","Breyanzi" = "Liso-cel"), levels = c("Liso-cel","Axi-cel") ),
    "Largest lesion" = Largest.lesion,
    "HCT-CI" = HCT.CI,
    "ALC pre-LD" = ALC.preLD,
    "LDH pre-LD" = LDH.preLD/1000,
    "Male sex" = ifelse(Sex == "M",1,0),
    CRS2.4=ifelse(CRS.grade == 0 | CRS.grade == 1 ,0,1),
    "Bridging category" = factor(recode(Bridging.response.category, "0" = "No bridging", "1" = "CR/PR", "2" = "SD/PD"), levels = c("No bridging","CR/PR","SD/PD"))
  )

data2 <- m.data %>%
  mutate(
    Product = factor(recode(CAR.T.Product.Type, "Yescarta" = "Axi-cel","Breyanzi" = "Liso-cel"), levels = c("Liso-cel","Axi-cel") ),
    "Largest lesion" = Largest.lesion,
    "HCT-CI" = HCT.CI,
    "ALC pre-LD" = ALC.preLD,
    "LDH pre-LD" = LDH.preLD/1000,
    "Male sex" = ifelse(Sex == "M",1,0),
    CRS2.4=ifelse(CRS.grade == 0 | CRS.grade == 1 ,0,1),
    "Bridging category" = factor(recode(Bridging.response.category, "0" = "No bridging", "1" = "CR/PR", "2" = "SD/PD"), levels = c("No bridging","CR/PR","SD/PD"))
  )

uv_tab <- tbl_uvregression(
  data0[c(preds)],
  method = lm,
  y = toci.doses
)

mv_tab<-lm( toci.doses ~ Product + `Largest lesion` + `ALC pre-LD` + Age + `Male sex` + `LDH pre-LD` + `Bridging category`, data0) %>% tbl_regression() 

uv_tab_IPTW <- tbl_uvregression(
  data1[c(preds)],
  method = lm,
  y = toci.doses,
  method.args = list(weights = data1$weights)
)

mv_tab_IPTW<-lm( toci.doses ~ Product + `Largest lesion` + `ALC pre-LD` + Age + `Male sex` + `LDH pre-LD` + `Bridging category`, data1, weights=weights ) %>%
  tbl_regression() 

uv_tab_PSM <- tbl_uvregression(
  data2[c(preds)],
  method = lm,
  y = toci.doses,
  method.args = list(weights = data2$weights)
)

mv_tab_PSM<-lm( toci.doses ~ Product + `Largest lesion` + `ALC pre-LD` + Age + `Male sex` + `LDH pre-LD` + `Bridging category`, data2, weights=weights ) %>%
  tbl_regression() 

row1 <- tbl_merge(list(uv_tab, mv_tab), tab_spanner = c("**Univariate**", "**Multivariable**"))
row2 <- tbl_merge(list(uv_tab_IPTW, mv_tab_IPTW))
row3 <- tbl_merge(list(uv_tab_PSM, mv_tab_PSM))
tbl_stack(list(row1, row2, row3), group_header = c("UNWEIGHTED ANALYSIS", "IPTW ANALYSIS", "PSM ANALYSIS"), quiet=TRUE)

Characteristic	Univariate				Multivariable
	N	Beta	95% CI1	p-value	Beta	95% CI1	p-value
UNWEIGHTED ANALYSIS
Product	149
Liso-cel		—	—		—	—
Axi-cel		0.58	0.21, 0.95	0.002	0.37	-0.05, 0.78	0.080
LDH pre-LD	148	0.17	0.02, 0.32	0.027	0.13	-0.02, 0.28	0.093
ALC pre-LD	149	-0.04	-0.14, 0.05	0.4	-0.04	-0.13, 0.06	0.4
Largest lesion	149	0.07	0.02, 0.12	0.004	0.06	0.01, 0.11	0.026
Age	149	-0.01	-0.03, 0.00	0.049	-0.01	-0.02, 0.01	0.5
Male sex	149	0.05	-0.32, 0.43	0.8	0.01	-0.36, 0.39	>0.9
CMV	149	1.1	0.77, 1.5	<0.001
Bridging category	145
No bridging		—	—		—	—
CR/PR		-0.50	-1.0, 0.03	0.064	-0.35	-0.88, 0.17	0.2
SD/PD		0.30	-0.13, 0.73	0.2	0.08	-0.38, 0.55	0.7
IPTW ANALYSIS
Product	126
Liso-cel		—	—		—	—
Axi-cel		0.36	-0.02, 0.73	0.062	0.25	-0.14, 0.65	0.2
LDH pre-LD	125	0.18	0.01, 0.35	0.036	0.14	-0.03, 0.32	0.10
ALC pre-LD	126	-0.04	-0.14, 0.06	0.5	-0.04	-0.14, 0.06	0.4
Largest lesion	126	0.06	0.01, 0.11	0.017	0.05	0.00, 0.10	0.049
Age	126	-0.01	-0.03, 0.00	0.10	-0.01	-0.02, 0.01	0.4
Male sex	126	-0.06	-0.43, 0.32	0.8	-0.07	-0.44, 0.31	0.7
CMV	126	1.1	0.77, 1.5	<0.001
Bridging category	123
No bridging		—	—		—	—
CR/PR		-0.49	-0.98, 0.01	0.053	-0.45	-0.94, 0.05	0.076
SD/PD		0.24	-0.18, 0.66	0.3	0.02	-0.45, 0.48	>0.9
PSM ANALYSIS
Product	149
Liso-cel		—	—		—	—
Axi-cel		0.40	0.08, 0.72	0.015	0.34	0.00, 0.68	0.052
LDH pre-LD	148	0.31	-0.07, 0.69	0.11	0.21	-0.19, 0.60	0.3
ALC pre-LD	149	0.00	-0.08, 0.08	>0.9	-0.01	-0.09, 0.08	0.8
Largest lesion	149	0.04	-0.01, 0.09	0.084	0.03	-0.03, 0.08	0.3
Age	149	-0.01	-0.03, 0.00	0.077	-0.01	-0.03, 0.01	0.2
Male sex	149	-0.07	-0.39, 0.25	0.7	-0.16	-0.48, 0.17	0.3
CMV	149	1.1	0.81, 1.4	<0.001
Bridging category	144
No bridging		—	—		—	—
CR/PR		-0.59	-1.0, -0.17	0.006	-0.51	-0.93, -0.08	0.020
SD/PD		0.23	-0.17, 0.62	0.3	0.10	-0.36, 0.55	0.7
1 CI = Confidence Interval

Cefepime

Show code

theme_gtsummary_compact()
preds <- c("cefepime.doses","Product", "LDH pre-LD","ALC pre-LD", "Largest lesion", "Age", "Male sex", "CMV", "Bridging category") 

data0 <- dataset %>%
  mutate(
    Product = factor(recode(CAR.T.Product.Type, "Yescarta" = "Axi-cel","Breyanzi" = "Liso-cel"), levels = c("Liso-cel","Axi-cel") ),
    "Largest lesion" = Largest.lesion,
    "HCT-CI" = HCT.CI,
    "ALC pre-LD" = ALC.preLD,
    "LDH pre-LD" = LDH.preLD/1000,
    "Male sex" = ifelse(Sex == "M",1,0),
    CRS2.4=ifelse(CRS.grade == 0 | CRS.grade == 1 ,0,1),
    "Bridging category" = factor(recode(Bridging.response.category, "0" = "No bridging", "1" = "CR/PR", "2" = "SD/PD"), levels = c("No bridging","CR/PR","SD/PD"))
  )

data1 <- IPTW.data %>%
  mutate(
    Product = factor(recode(CAR.T.Product.Type, "Yescarta" = "Axi-cel","Breyanzi" = "Liso-cel"), levels = c("Liso-cel","Axi-cel") ),
    "Largest lesion" = Largest.lesion,
    "HCT-CI" = HCT.CI,
    "ALC pre-LD" = ALC.preLD,
    "LDH pre-LD" = LDH.preLD/1000,
    "Male sex" = ifelse(Sex == "M",1,0),
    CRS2.4=ifelse(CRS.grade == 0 | CRS.grade == 1 ,0,1),
    "Bridging category" = factor(recode(Bridging.response.category, "0" = "No bridging", "1" = "CR/PR", "2" = "SD/PD"), levels = c("No bridging","CR/PR","SD/PD"))
  )

data2 <- m.data %>%
  mutate(
    Product = factor(recode(CAR.T.Product.Type, "Yescarta" = "Axi-cel","Breyanzi" = "Liso-cel"), levels = c("Liso-cel","Axi-cel") ),
    "Largest lesion" = Largest.lesion,
    "HCT-CI" = HCT.CI,
    "ALC pre-LD" = ALC.preLD,
    "LDH pre-LD" = LDH.preLD/1000,
    "Male sex" = ifelse(Sex == "M",1,0),
    CRS2.4=ifelse(CRS.grade == 0 | CRS.grade == 1 ,0,1),
    "Bridging category" = factor(recode(Bridging.response.category, "0" = "No bridging", "1" = "CR/PR", "2" = "SD/PD"), levels = c("No bridging","CR/PR","SD/PD"))
  )

uv_tab <- tbl_uvregression(
  data0[c(preds)],
  method = lm,
  y = cefepime.doses
)

mv_tab<-lm( cefepime.doses ~ Product + `Largest lesion` + `ALC pre-LD` + Age + `Male sex` + `LDH pre-LD` + `Bridging category`, data0) %>% tbl_regression() 

uv_tab_IPTW <- tbl_uvregression(
  data1[c(preds)],
  method = lm,
  y = cefepime.doses,
  method.args = list(weights = data1$weights)
)

mv_tab_IPTW<-lm( cefepime.doses ~ Product + `Largest lesion` + `ALC pre-LD` + Age + `Male sex` + `LDH pre-LD` + `Bridging category`, data1, weights=weights ) %>%
  tbl_regression() 

uv_tab_PSM <- tbl_uvregression(
  data2[c(preds)],
  method = lm,
  y = cefepime.doses,
  method.args = list(weights = data2$weights)
)

mv_tab_PSM<-lm( cefepime.doses ~ Product + `Largest lesion` + `ALC pre-LD` + Age + `Male sex` + `LDH pre-LD` + `Bridging category`, data2, weights=weights ) %>%
  tbl_regression() 

row1 <- tbl_merge(list(uv_tab, mv_tab), tab_spanner = c("**Univariate**", "**Multivariable**"))
row2 <- tbl_merge(list(uv_tab_IPTW, mv_tab_IPTW))
row3 <- tbl_merge(list(uv_tab_PSM, mv_tab_PSM))
tbl_stack(list(row1, row2, row3), group_header = c("UNWEIGHTED ANALYSIS", "IPTW ANALYSIS", "PSM ANALYSIS"), quiet=TRUE)

Characteristic	Univariate				Multivariable
	N	Beta	95% CI1	p-value	Beta	95% CI1	p-value
UNWEIGHTED ANALYSIS
Product	149
Liso-cel		—	—		—	—
Axi-cel		4.9	2.0, 7.8	<0.001	4.8	1.5, 8.0	0.005
LDH pre-LD	148	1.2	0.02, 2.4	0.046	0.95	-0.24, 2.2	0.12
ALC pre-LD	149	0.58	-0.17, 1.3	0.13	0.79	0.03, 1.5	0.041
Largest lesion	149	0.19	-0.21, 0.60	0.3	0.06	-0.36, 0.49	0.8
Age	149	-0.11	-0.21, 0.00	0.056	-0.06	-0.17, 0.06	0.3
Male sex	149	1.2	-1.8, 4.1	0.4	1.7	-1.3, 4.6	0.3
CMV	149	3.4	0.26, 6.5	0.034
Bridging category	145
No bridging		—	—		—	—
CR/PR		0.31	-3.9, 4.6	0.9	2.0	-2.1, 6.2	0.3
SD/PD		1.2	-2.2, 4.7	0.5	0.99	-2.7, 4.6	0.6
IPTW ANALYSIS
Product	126
Liso-cel		—	—		—	—
Axi-cel		4.0	1.1, 7.0	0.008	3.9	0.72, 7.1	0.017
LDH pre-LD	125	1.3	-0.03, 2.7	0.056	1.1	-0.28, 2.5	0.12
ALC pre-LD	126	0.62	-0.19, 1.4	0.13	0.78	-0.03, 1.6	0.060
Largest lesion	126	0.08	-0.32, 0.47	0.7	0.06	-0.37, 0.48	0.8
Age	126	-0.08	-0.19, 0.04	0.2	-0.07	-0.19, 0.05	0.2
Male sex	126	0.28	-2.7, 3.3	0.9	0.94	-2.1, 4.0	0.5
CMV	126	3.4	0.29, 6.4	0.032
Bridging category	123
No bridging		—	—		—	—
CR/PR		0.09	-4.0, 4.1	>0.9	1.0	-3.0, 5.0	0.6
SD/PD		-0.04	-3.5, 3.4	>0.9	-0.27	-4.0, 3.5	0.9
PSM ANALYSIS
Product	149
Liso-cel		—	—		—	—
Axi-cel		3.8	1.0, 6.6	0.008	4.0	0.96, 7.0	0.010
LDH pre-LD	148	3.8	0.44, 7.1	0.027	2.8	-0.64, 6.3	0.11
ALC pre-LD	149	0.71	0.00, 1.4	0.051	0.83	0.09, 1.6	0.028
Largest lesion	149	0.16	-0.27, 0.59	0.5	-0.01	-0.47, 0.45	>0.9
Age	149	-0.11	-0.25, 0.03	0.14	-0.11	-0.26, 0.04	0.2
Male sex	149	1.1	-1.7, 3.9	0.4	1.2	-1.7, 4.0	0.4
CMV	149	4.9	1.8, 8.0	0.002
Bridging category	144
No bridging		—	—		—	—
CR/PR		-1.5	-5.3, 2.3	0.4	-0.85	-4.6, 2.9	0.7
SD/PD		0.45	-3.1, 4.0	0.8	-0.44	-4.4, 3.6	0.8
1 CI = Confidence Interval

PROBABILITY PLOTS

Response based on labs and imaging

Probability of CR: LDH pre-LD, stratified by product

Show code

data1 <- filter(dataset, Infused == 1) %>%
  mutate(
    CAR.T.Product.Type = recode(CAR.T.Product.Type, "Breyanzi"="Liso-cel","Yescarta"="Axi-cel")
  )

dd <- datadist(data1)
options(datadist='dd')

fit <- lrm(Best.CR ~ CAR.T.Product.Type * LDH.preLD, data=data1)

data.frame(Predict(fit,LDH.preLD,CAR.T.Product.Type,fun = plogis)) %>%
  ggplot() +
  geom_line(aes(x=LDH.preLD,y=yhat,col=CAR.T.Product.Type)) +
  geom_ribbon(aes(x=LDH.preLD,ymin=lower, ymax=upper,fill=CAR.T.Product.Type),alpha=0.10)+
  labs(x = "LDH pre-LD (U/L)",
       y = "Probability of CR")+
  #scale_x_continuous(labels = alc_breaks,breaks = c(-.5,-.25,0,.25))+
  #ggtitle("Figure.")+
  theme_bw() +
  theme(title = element_text(size=16,face="bold"),legend.title=element_blank()) + 
  ylim(0,1)

Probability of CR: largest lesion, stratified by product

Show code

data1 <- filter(dataset, Infused == 1) %>%
  mutate(
    CAR.T.Product.Type = recode(CAR.T.Product.Type, "Breyanzi"="Liso-cel","Yescarta"="Axi-cel")
  )

dd <- datadist(data1)
options(datadist='dd')

fit <- lrm(Best.CR ~ CAR.T.Product.Type*Largest.lesion, data=data1)

data.frame(Predict(fit,Largest.lesion,CAR.T.Product.Type,fun = plogis)) %>%
  ggplot() +
  geom_line(aes(x=Largest.lesion,y=yhat,col=CAR.T.Product.Type)) +
  geom_ribbon(aes(x=Largest.lesion,ymin=lower, ymax=upper,fill=CAR.T.Product.Type),alpha=0.10)+
  labs(x = "Largest lesion (cm)",
       y = "Probability of CR")+
  #scale_x_continuous(labels = alc_breaks,breaks = c(-.5,-.25,0,.25))+
  #ggtitle("Figure.")+
  theme_bw() +
  theme(title = element_text(size=16,face="bold"),legend.title=element_blank()) + 
  ylim(0,1)

Probability of CR and PD: ALC (RCS)

Show code

data1 <- filter(dataset, Infused == 1)

dd <- datadist(data1)
options(datadist='dd')

fit <- lrm(Best.CR ~ rcs(ALC.preLD,3), data=data1)
df1 <- data.frame(Predict(fit,ALC.preLD,fun = plogis))

fit2 <- lrm(Best.PD ~ rcs(ALC.preLD,3), data=data1)
df2 <- data.frame(Predict(fit2,ALC.preLD,fun = plogis))

ggplot() + 
  geom_line(data=df1, aes(x=ALC.preLD,y=yhat, color = "CR")) + geom_ribbon( data=df1, aes(x=ALC.preLD,ymin=lower, ymax=upper),alpha=0.10)+
  geom_line(data=df2, aes(x=ALC.preLD,y=yhat, color ="PD" )) + geom_ribbon(data = df2, aes(x=ALC.preLD,ymin=lower, ymax=upper),alpha=0.10)+
  labs(x = "ALC pre-LD (10\U00B3/\U00B5L)",
       y = "Probability")+
  theme_bw() +
  scale_y_continuous(breaks=c(0,0.2,0.4,0.6,0.8,1.0), limits=c(0, 1)) + 
  scale_x_continuous(breaks=c(0.25,0.5,0.75,1.0,1.25,1.5,1.75), limits=c(0.25, 1.75)) + 
  theme(legend.position = c(0, 1),legend.justification = c(-9, 1.1))+
  scale_color_manual(values = c ("steelblue","coral2")) + 
  theme(title = element_text(size=20,face="bold"),legend.title=element_blank())

Hazard of PFS: LDH pre-LD

Show code

data1 <- filter(dataset, Infused == 1)

S <- Surv(data1$Days.to.relapse.death.or.DLC, data1$Relapse.or.death)
f <- cph(S ~ LDH.preLD * CAR.T.Product.Type, x=TRUE, y=TRUE,surv=TRUE,data=data1)

model <- Predict(f, CAR.T.Product.Type, LDH.preLD, fun=function(x) exp(x) )

ggplot(as.data.frame(model),aes(x=LDH.preLD, y=yhat, z = CAR.T.Product.Type)) + 
  geom_line(
    aes(col = CAR.T.Product.Type)
  ) + 
  geom_ribbon(data = model, aes(ymin=lower, ymax=upper), alpha=0.2, linetype=0) + 
  theme_bw() + 
  labs(
    #title = "Survival by product and LDH pre-LD",
    x = "LDH pre-LD (U/L)",
    y = "PFS (hazard)") +
  theme(title = element_text(size=16,face="bold"),legend.title=element_blank())

Probability of CR and PD: LDH pre-LD

Show code

data1 <- filter(dataset, Infused == 1)

dd <- datadist(data1)
options(datadist='dd')

fit <- lrm(Best.CR ~ LDH.preLD, data=data1)

df1 <- data.frame(Predict(fit,LDH.preLD,fun = plogis))



fit2 <- lrm(Best.PD ~ LDH.preLD, data=data1)
df2 <- data.frame(Predict(fit2,LDH.preLD,fun = plogis))

ggplot() + 
  geom_line(data=df1, aes(x=LDH.preLD,y=yhat, color = "CR")) + geom_ribbon( data=df1, aes(x=LDH.preLD,ymin=lower, ymax=upper),alpha=0.10)+
  geom_line(data=df2, aes(x=LDH.preLD,y=yhat, color ="PD" )) + geom_ribbon(data = df2, aes(x=LDH.preLD,ymin=lower, ymax=upper),alpha=0.10)+
  labs(x = "LDH pre-LD (U/L)",
       y = "Probability")+
  theme_classic() +
  scale_y_continuous(breaks=c(0,0.2,0.4,0.6,0.8,1.0), limits=c(0, 1)) + 
  #scale_x_continuous(breaks=c(0.25,0.5,0.75,1.0,1.25,1.5,1.75), limits=c(0.25, 1.75)) + 
  theme(legend.position = c(0, 1),legend.justification = c(-9, 1.1))+
  scale_color_manual(values = c ("steelblue","coral2")) + 
  theme(title = element_text(size=20,face="bold"),legend.title=element_blank())

Probability of CR: LDH pre-LD

Show code

data1 <- filter(dataset, Infused == 1)

dd <- datadist(data1)
options(datadist='dd')

fit <- lrm(Best.CR ~ rcs(LDH.preLD,3), data=data1)

data.frame(Predict(fit,LDH.preLD,fun = plogis)) %>%
  ggplot() +
  geom_line(aes(x=LDH.preLD,y=yhat)) +
  geom_ribbon(aes(x=LDH.preLD,ymin=lower, ymax=upper),alpha=0.10)+
  labs(x = "LDH pre-LD (U/L)",
       y = "Probability of CR")+
  #scale_x_continuous(labels = alc_breaks,breaks = c(-.5,-.25,0,.25))+
  #ggtitle("Figure.")+
  ylim(0,1) +
  theme_classic() +
  theme(title = element_text(size=16,face="bold"),legend.title=element_blank())

Probability of CR: largest Lesion size, stratified by product

Show code

data1 <- filter(dataset, Infused == 1)

dd <- datadist(data1)
options(datadist='dd')

fit <- lrm(Best.CR ~ CAR.T.Product.Type * rcs(Largest.lesion,3), data=data1)

data.frame(Predict(fit,Largest.lesion,CAR.T.Product.Type,fun = plogis)) %>%
  ggplot() +
  geom_line(aes(x=Largest.lesion ,y=yhat,col=CAR.T.Product.Type)) +
  geom_ribbon(aes(x=Largest.lesion ,ymin=lower, ymax=upper,fill=CAR.T.Product.Type),alpha=0.10)+
  labs(x = "Largest lesion (cm)",
       y = "Probability of CR")+
  theme_bw() +
  theme(title = element_text(size=16,face="bold"),legend.title=element_blank())

Probability of CR: Largest Lesion size (RCS)

Show code

data1 <- filter(dataset, Infused == 1)

dd <- datadist(data1)
options(datadist='dd')

fit <- lrm(Best.CR ~ rcs(Largest.lesion,3), data=data1)

data.frame(Predict(fit,Largest.lesion,fun = plogis)) %>%
  ggplot() +
  geom_line(aes(x=Largest.lesion ,y=yhat)) +
  geom_ribbon(aes(x=Largest.lesion ,ymin=lower, ymax=upper),alpha=0.10)+
  labs(x = "Largest lesion (cm)",
       y = "Probability of CR")+
  ylim(0,1) +
  theme_classic() +
  theme(title = element_text(size=16,face="bold"),legend.title=element_blank())

Ferritin!

Probability of ICANS based on day 3 ferritin

Show code

data1 <- filter(dataset, Infused == 1) %>%
  mutate(
    ICANS0 = ifelse( ICANS.grade==0,1,0),
    ICANS1.2 = ifelse( ICANS.grade==1 | ICANS.grade == 2,1,0),
    ICANS.any.grade = ifelse( ICANS.grade != 0 ,1,0),
    ICANS3.4 = ifelse( ICANS.grade==3 | ICANS.grade == 4,1,0),
    ferritin.ratio = log(ferritin.d3 / ferritin.d0),
    ferritin.d3 = log10(ferritin.d3)
  )

dd <- datadist(data1)
options(datadist='dd')

fit <- lrm(ICANS0 ~ ferritin.d3 + CAR.T.Product.Type, data=data1)
df1 <- data.frame(Predict(fit,ferritin.d3,fun = plogis))


fit2 <- lrm(ICANS1.2 ~ ferritin.d3 + CAR.T.Product.Type, data=data1)
df2 <- data.frame(Predict(fit2,ferritin.d3,fun = plogis))

fit3 <- lrm(ICANS3.4 ~ ferritin.d3 + CAR.T.Product.Type, data=data1)
df3 <- data.frame(Predict(fit3,ferritin.d3,fun = plogis))

ggplot() + 
  geom_line(data=df1, aes(x=ferritin.d3,y=yhat, color = "G0 ICANS")) + geom_ribbon( data=df1, aes(x=ferritin.d3,ymin=lower, ymax=upper),alpha=0.10)+
  geom_line(data=df2, aes(x=ferritin.d3,y=yhat, color ="G1-2 ICANS" )) + geom_ribbon(data = df2, aes(x=ferritin.d3,ymin=lower, ymax=upper),alpha=0.10)+
    geom_line(data=df3, aes(x=ferritin.d3,y=yhat, color ="G\u22653 ICANS" )) + geom_ribbon(data = df3, aes(x=ferritin.d3,ymin=lower, ymax=upper),alpha=0.10)+
  labs(x = "log(ferritin, day 3) (ng/mL)",
       y = "Probability")+
  theme_bw() +
  scale_y_continuous(breaks=c(0,0.2,0.4,0.6,0.8,1.0), limits=c(0, 1)) + 
  #theme(legend.position = c(0, 1),legend.justification = c(-4.5, 1.1))+
  scale_color_manual(values = c ("steelblue","coral2","purple")) + 
  theme(title = element_text(size=20,face="bold"),legend.title=element_blank())

Probability of CRS based on day 3 ferritin

Show code

data1 <- filter(dataset, Infused == 1) %>%
  mutate(
    ICANS0 = ifelse( ICANS.grade==0,1,0),
    ICANS1.2 = ifelse( ICANS.grade==1 | ICANS.grade == 2,1,0),
    ICANS.any.grade = ifelse( ICANS.grade != 0 ,1,0),
    ICANS3.4 = ifelse( ICANS.grade==3 | ICANS.grade == 4,1,0),
    CRS0 = ifelse( CRS.grade==0,1,0),
    CRS1.2 = ifelse( CRS.grade==1 | CRS.grade == 2,1,0),
    CRS.any.grade = ifelse( CRS.grade != 0 ,1,0),
    CRS3.4 = ifelse( CRS.grade==3 | CRS.grade == 4,1,0),
    
    ferritin.ratio = log(ferritin.d3 / ferritin.d0),
    ferritin.d3 = log10(ferritin.d3)
  )

dd <- datadist(data1)
options(datadist='dd')

fit <- lrm(CRS0 ~ ferritin.d3 + CAR.T.Product.Type, data=data1)
df1 <- data.frame(Predict(fit,ferritin.d3,fun = plogis))


fit2 <- lrm(CRS1.2 ~ ferritin.d3 + CAR.T.Product.Type, data=data1)
df2 <- data.frame(Predict(fit2,ferritin.d3,fun = plogis))

fit3 <- lrm(CRS3.4 ~ ferritin.d3 + CAR.T.Product.Type, data=data1)
df3 <- data.frame(Predict(fit3,ferritin.d3,fun = plogis))

ggplot() + 
  geom_line(data=df1, aes(x=ferritin.d3,y=yhat, color = "G0 CRS")) + geom_ribbon( data=df1, aes(x=ferritin.d3,ymin=lower, ymax=upper),alpha=0.10)+
  geom_line(data=df2, aes(x=ferritin.d3,y=yhat, color ="G1-2 CRS" )) + geom_ribbon(data = df2, aes(x=ferritin.d3,ymin=lower, ymax=upper),alpha=0.10)+
    geom_line(data=df3, aes(x=ferritin.d3,y=yhat, color ="G\u22653 CRS" )) + geom_ribbon(data = df3, aes(x=ferritin.d3,ymin=lower, ymax=upper),alpha=0.10)+
  labs(x = "log(ferritin, day 3) (ng/mL)",
       y = "Probability")+
  theme_bw() +
  scale_y_continuous(breaks=c(0,0.2,0.4,0.6,0.8,1.0), limits=c(0, 1)) + 
  #theme(legend.position = c(0, 1),legend.justification = c(-4.5, 1.1))+
  scale_color_manual(values = c ("steelblue","coral2","purple")) + 
  theme(title = element_text(size=20,face="bold"),legend.title=element_blank())

Probability of RESPONSE based on day 3 ferritin

Show code

data1 <- filter(dataset, Infused == 1) %>%
  mutate(
    ICANS0 = ifelse( ICANS.grade==0,1,0),
    ICANS1.2 = ifelse( ICANS.grade==1 | ICANS.grade == 2,1,0),
    ICANS.any.grade = ifelse( ICANS.grade != 0 ,1,0),
    ICANS3.4 = ifelse( ICANS.grade==3 | ICANS.grade == 4,1,0),
    CRS0 = ifelse( CRS.grade==0,1,0),
    CRS1.2 = ifelse( CRS.grade==1 | CRS.grade == 2,1,0),
    CRS.any.grade = ifelse( CRS.grade != 0 ,1,0),
    CRS3.4 = ifelse( CRS.grade==3 | CRS.grade == 4,1,0),
    
    ferritin.ratio = log(ferritin.d3 / ferritin.d0),
    ferritin.d3 = log10(ferritin.d3)
  )

dd <- datadist(data1)
options(datadist='dd')

fit <- lrm(Best.CR ~ ferritin.d3 + CAR.T.Product.Type, data=data1)
df1 <- data.frame(Predict(fit,ferritin.d3,fun = plogis))


fit2 <- lrm(Best.PR ~ ferritin.d3 + CAR.T.Product.Type, data=data1)
df2 <- data.frame(Predict(fit2,ferritin.d3,fun = plogis))

fit3 <- lrm(Best.PD ~ ferritin.d3 + CAR.T.Product.Type, data=data1)
df3 <- data.frame(Predict(fit3,ferritin.d3,fun = plogis))

ggplot() + 
  geom_line(data=df1, aes(x=ferritin.d3,y=yhat, color = "CR")) + geom_ribbon( data=df1, aes(x=ferritin.d3,ymin=lower, ymax=upper),alpha=0.10)+
  geom_line(data=df2, aes(x=ferritin.d3,y=yhat, color ="PR" )) + geom_ribbon(data = df2, aes(x=ferritin.d3,ymin=lower, ymax=upper),alpha=0.10)+
    geom_line(data=df3, aes(x=ferritin.d3,y=yhat, color ="PD" )) + geom_ribbon(data = df3, aes(x=ferritin.d3,ymin=lower, ymax=upper),alpha=0.10)+
  labs(x = "log(ferritin, day 3) (ng/mL)",
       y = "Probability")+
  theme_bw() +
  scale_y_continuous(breaks=c(0,0.2,0.4,0.6,0.8,1.0), limits=c(0, 1)) + 
  #theme(legend.position = c(0, 1),legend.justification = c(-8.5, 1.1))+
  scale_color_manual(values = c ("steelblue","coral2","purple")) + 
  theme(title = element_text(size=20,face="bold"),legend.title=element_blank())

Probability of CR or PD based on day 3 ferritin - RESTRICTED CUBIC SPLINES model

Show code

data1 <- filter(dataset, Infused == 1) %>%
  mutate(
    ICANS0.2 = ifelse( as.numeric(ICANS.grade)<=2,1,0),
    ICANS.any.grade = ifelse( ICANS.grade != 0 ,1,0),
    ICANS3.4 = ifelse( ICANS.grade ==3 | ICANS.grade == 4,1,0),
    ferritin.ratio = ferritin.d3 / ferritin.d0
    #ferritin.d3 = log(ferritin.d3)
  )

dd <- datadist(data1)
options(datadist='dd')

fit <- lrm(Best.CR ~ rcs(ferritin.d3,3), data=data1)

df1 <- data.frame(Predict(fit,ferritin.d3,fun = plogis))



fit2 <- lrm(Best.PD ~ rcs(ferritin.d3,3), data=data1)
df2 <- data.frame(Predict(fit2,ferritin.d3,fun = plogis))

ggplot() + 
  geom_line(data=df1, aes(x=ferritin.d3,y=yhat, color = "CR")) + geom_ribbon( data=df1, aes(x=ferritin.d3,ymin=lower, ymax=upper),alpha=0.10)+
  geom_line(data=df2, aes(x=ferritin.d3,y=yhat, color ="PD" )) + geom_ribbon(data = df2, aes(x=ferritin.d3,ymin=lower, ymax=upper),alpha=0.10)+
  labs(x = "Ferritin, day 3 (ng/mL)",
       y = "Probability")+
  theme_bw() +
  scale_y_continuous(breaks=c(0,0.2,0.4,0.6,0.8,1.0), limits=c(0, 1)) + 
  #scale_x_continuous(breaks=c(0.25,0.5,0.75,1.0,1.25,1.5,1.75), limits=c(0.25, 1.75)) + 
  #theme(legend.position = c(0, 1),legend.justification = c(-9, 1.1))+
  scale_color_manual(values = c ("steelblue","coral2")) + 
  theme(title = element_text(size=20,face="bold"),legend.title=element_blank())

Hazard of OS and PFS: D3 ferritin

Show code

data1 <- filter(dataset, Infused == 1) %>%
  mutate(
    ferritin.d3 = log10(ferritin.d3)
  )

dd <- datadist(data1)
options(datadist='dd')

S1 <- Surv(data1$Days.to.DLC, data1$Death)
f1 <- cph(S1 ~ ferritin.d3 , x=TRUE, y=TRUE,surv=TRUE,data=data1)
model1 <- Predict(f1, ferritin.d3, fun=function(x) exp(x) )


S2 <- Surv(data1$Days.to.relapse.death.or.DLC, data1$Relapse.or.death)
f2 <- cph(S2 ~ ferritin.d3 , x=TRUE, y=TRUE,surv=TRUE,data=data1)
model2 <- Predict(f2, ferritin.d3, fun=function(x) exp(x) )

ggplot() + 
  geom_line(data=model1, aes(x=ferritin.d3, y=yhat, color = "OS")) +  geom_ribbon(data = model1, aes(x=ferritin.d3, ymin=lower, ymax=upper), alpha=0.1) + 
  geom_line(data=model2, aes(x=ferritin.d3, y=yhat, color = "PFS")) +  geom_ribbon(data = model2, aes(x=ferritin.d3, ymin=lower, ymax=upper), alpha=0.1) +
  theme_bw() + 
  labs(
    #title = "Survival by product and ferritin at day 3",
    x = "log(ferritin, day 3)",
    y = "Hazard") +
  #theme(legend.position = c(0, 1),legend.justification = c(-0.8, 1.2))+
  scale_color_manual(values = c ("coral2","steelblue")) + 
  theme(title = element_text(size=20,face="bold"),legend.title=element_blank())

Hazard of PFS: D3 ferritin

Show code

data1 <- filter(dataset, Infused == 1) %>%
  mutate(
    ferritin.d3 = log10(ferritin.d3)
  )

dd <- datadist(data1)
options(datadist='dd')

S <- Surv(data1$Days.to.relapse.death.or.DLC, data1$Relapse.or.death)
f <- cph(S ~ ferritin.d3 , x=TRUE, y=TRUE,surv=TRUE,data=data1)

model <- Predict(f, ferritin.d3, fun=function(x) exp(x) )

ggplot(as.data.frame(model),aes(x=ferritin.d3, y=yhat)) + 
  geom_line(
    aes()
  ) + 
  geom_ribbon(data = model, aes(ymin=lower, ymax=upper), alpha=0.2, linetype=0) + 
  theme_bw() + 
  labs(
    #title = "Survival by product and ferritin at day 3",
    x = "log(ferritin), day 3",
    y = "PFS (hazard)") +
  theme(title = element_text(size=16,face="bold"),legend.title=element_blank())

Hazard of PFS: D3 ferritin - stratified by CAR-T-product

Show code

data1 <- filter(dataset, Infused == 1) %>%
  mutate(
    ferritin.d3 = log10(ferritin.d3)
  )

dd <- datadist(data1)
options(datadist='dd')

S <- Surv(data1$Days.to.relapse.death.or.DLC, data1$Relapse.or.death)
f <- cph(S ~ ferritin.d3 + CAR.T.Product.Type , x=TRUE, y=TRUE,surv=TRUE,data=data1)

model <- Predict(f, CAR.T.Product.Type, ferritin.d3, fun=function(x) exp(x) )

ggplot(as.data.frame(model),aes(x=ferritin.d3, y=yhat, z = CAR.T.Product.Type)) + 
  geom_line(
    aes(col = CAR.T.Product.Type)
  ) + 
  geom_ribbon(data = model, aes(ymin=lower, ymax=upper), alpha=0.2, linetype=0) + 
  theme_bw() + 
  labs(
    #title = "Survival by product and ferritin at day 3",
    x = "log(ferritin), day 3",
    y = "PFS (hazard)") +
  theme(title = element_text(size=16,face="bold"),legend.title=element_blank())

ICANS/CRS

Probability of Grade 0-2 and Grade 3-4 ICANS: LDH pre-LD (RCS)

Show code

data1 <- filter(dataset, Infused == 1) %>%
  mutate(
    ICANS0.2 = ifelse( ICANS.grade == 2 | ICANS.grade == 1 | ICANS.grade == 0,1,0),
    ICANS3.4 = ifelse( ICANS.grade == 3 | ICANS.grade == 4,1,0)
    #ferritin.d3 = log(ferritin.d3)
  )

dd <- datadist(data1)
options(datadist='dd')

fit <- lrm(ICANS0.2 ~ rcs(LDH.preLD,3), data=data1)

df1 <- data.frame(Predict(fit,LDH.preLD,fun = plogis))



fit2 <- lrm(ICANS3.4 ~ rcs(LDH.preLD,3), data=data1)
df2 <- data.frame(Predict(fit2,LDH.preLD,fun = plogis))

ggplot() + 
  geom_line(data=df1, aes(x=LDH.preLD,y=yhat, color = "0-2")) + geom_ribbon( data=df1, aes(x=LDH.preLD,ymin=lower, ymax=upper),alpha=0.10)+
  geom_line(data=df2, aes(x=LDH.preLD,y=yhat, color ="3-4" )) + geom_ribbon(data = df2, aes(x=LDH.preLD,ymin=lower, ymax=upper),alpha=0.10)+
  labs(x = "LDH pre-LD (U/L)",
       y = "Probability")+
  theme_bw() +
  scale_y_continuous(breaks=c(0,0.2,0.4,0.6,0.8,1.0), limits=c(0, 1)) + 
  #scale_x_continuous(breaks=c(0.25,0.5,0.75,1.0,1.25,1.5,1.75), limits=c(0.25, 1.75)) + 
  #theme(legend.position = c(0, 1),legend.justification = c(-9, 1.1))+
  scale_color_manual(values = c ("steelblue","coral2")) + 
  theme(title = element_text(size=20,face="bold"),legend.title=element_blank())

Probability of any grade CRS based on Age

Show code

data1 <- filter(dataset, Infused == 1) %>%
  select(CRS.grade, ICANS.grade, Age, CAR.T.Product.Type) %>%
  mutate(
    CRS.any = ifelse( CRS.grade != 0,1,0),
    ICANS.any = ifelse( ICANS.grade !=0,1,0)
  )

dd <- datadist(data1)
options(datadist='dd')

fit <- lrm(CRS.any ~ CAR.T.Product.Type *Age, data=data1)

data.frame(Predict(fit,Age,CAR.T.Product.Type,fun = plogis)) %>%
  ggplot() +
  geom_line(aes(x=Age,y=yhat,col=CAR.T.Product.Type)) +
  geom_ribbon(aes(x=Age,ymin=lower, ymax=upper,fill=CAR.T.Product.Type),alpha=0.10)+
  labs(x = "Age (years)",
       y = "Probability of any grade CRS")+
  theme_bw() +
  theme(title = element_text(size=16,face="bold"),legend.title=element_blank())

Probability of any grade ICANS based on Age

Show code

data1 <- filter(dataset, Infused == 1) %>%
  select(CRS.grade, ICANS.grade, Age, CAR.T.Product.Type) %>%
  mutate(
    CRS.any = ifelse( CRS.grade != 0,1,0),
    ICANS.any = ifelse( ICANS.grade !=0,1,0)
  )
dd <- datadist(data1)
options(datadist='dd')

dd <- datadist(data1)
options(datadist='dd')

fit <- lrm(ICANS.any ~ CAR.T.Product.Type * Age, data=data1)

data.frame(Predict(fit,Age,CAR.T.Product.Type,fun = plogis)) %>%
  ggplot() +
  geom_line(aes(x=Age,y=yhat,col=CAR.T.Product.Type)) +
  geom_ribbon(aes(x=Age,ymin=lower, ymax=upper,fill=CAR.T.Product.Type),alpha=0.10)+
  labs(x = "Age (years)",
       y = "Probability of any grade ICANS")+
  #scale_x_continuous(labels = alc_breaks,breaks = c(-.5,-.25,0,.25))+
  #ggtitle("Figure.")+
  theme_bw() +
  theme(title = element_text(size=16,face="bold"),legend.title=element_blank())

Probability of Any grade and G3+ CRS: Age

Show code

data1 <- filter(dataset, Infused == 1) %>%
  select(CRS.grade, ICANS.grade, Age, CAR.T.Product.Type) %>%
  mutate(
    CRS3.4 = ifelse( CRS.grade == 3 | CRS.grade == 4,1,0),
    ICANS3.4 = ifelse( ICANS.grade == 3 | ICANS.grade == 4,1,0),
    CRS.any = ifelse( CRS.grade !=0 ,1,0),
    ICANS.any = ifelse( ICANS.grade !=0,1,0)
  )

dd <- datadist(data1)
options(datadist='dd')


fit1 <- lrm(CRS3.4 ~ CAR.T.Product.Type + Age, data=data1)

df1 <- data.frame(Predict(fit1,Age,fun = plogis))



fit2 <- lrm(CRS.any ~ CAR.T.Product.Type + Age, data=data1)
df2 <- data.frame(Predict(fit2,Age,fun = plogis))

ggplot() + 
  geom_line(data=df1, aes(x=Age,y=yhat, color = "\u22653")) + geom_ribbon( data=df1, aes(x=Age,ymin=lower, ymax=upper),alpha=0.10)+
  geom_line(data=df2, aes(x=Age,y=yhat, color ="Any" )) + geom_ribbon(data = df2, aes(x=Age,ymin=lower, ymax=upper),alpha=0.10)+
  labs(x = "Age (years)",
       y = "Probability of CRS")+
  theme_bw() +
  scale_y_continuous(breaks=c(0,0.2,0.4,0.6,0.8,1.0), limits=c(0, 1)) + 
  #scale_x_continuous(breaks=c(0.25,0.5,0.75,1.0,1.25,1.5,1.75), limits=c(0.25, 1.75)) + 
  #theme(legend.position = c(0, 1),legend.justification = c(-8.3, 3))+
  scale_color_manual(values = c ("coral2","steelblue")) + 
  theme(title = element_text(size=20,face="bold"),legend.title=element_blank())

Probability of any grade and 3+ ICANS: Age

Show code

data1 <- filter(dataset, Infused == 1) %>%
  select(CRS.grade, ICANS.grade, Age, CAR.T.Product.Type) %>%
  mutate(
    CRS3.4 = ifelse( CRS.grade == 3 | CRS.grade == 4,1,0),
    ICANS3.4 = ifelse( ICANS.grade == 3 | ICANS.grade == 4,1,0),
    CRS.any = ifelse( CRS.grade !=0 ,1,0),
    ICANS.any = ifelse( ICANS.grade !=0,1,0)
  )

dd <- datadist(data1)
options(datadist='dd')

fit1 <- lrm(ICANS3.4 ~ CAR.T.Product.Type + Age, data=data1)
df1 <- data.frame(Predict(fit1,Age,fun = plogis))

fit2 <- lrm(ICANS.any ~ CAR.T.Product.Type + Age, data=data1)
df2 <- data.frame(Predict(fit2,Age,fun = plogis))

ggplot() + 
  geom_line(data=df1, aes(x=Age,y=yhat, color = "\u22653")) + geom_ribbon( data=df1, aes(x=Age,ymin=lower, ymax=upper),alpha=0.10)+
  geom_line(data=df2, aes(x=Age,y=yhat, color ="Any" )) + geom_ribbon(data = df2, aes(x=Age,ymin=lower, ymax=upper),alpha=0.10)+
  labs(x = "Age (years)",
       y = "Probability of ICANS")+
  theme_bw() +
  scale_y_continuous(breaks=c(0,0.2,0.4,0.6,0.8,1.0), limits=c(0, 1)) + 
  #scale_x_continuous(breaks=c(0.25,0.5,0.75,1.0,1.25,1.5,1.75), limits=c(0.25, 1.75)) + 
  #theme(legend.position = c(0, 1),legend.justification = c(-8.3, 1.1))+
  scale_color_manual(values = c ("coral2","steelblue")) + 
  theme(title = element_text(size=20,face="bold"),legend.title=element_blank())

PFS & OS

Hazard of PFS: LDH - stratified by bulky status

Show code

data1 <- filter(dataset, Infused == 1) %>%
  mutate(
    Bulky = recode(Bulky, "1"="Bulky","0"="Not bulky")
  )

dd <- datadist(data1)
options(datadist='dd')

S <- Surv(data1$Days.to.relapse.death.or.DLC, data1$Relapse.or.death)
f <- cph(S ~ LDH.preLD * Bulky , x=TRUE, y=TRUE,surv=TRUE,data=data1)

med <- Quantile(f)
model <- Predict(f, Bulky, LDH.preLD, fun=function(x) exp(x) )

ggplot(as.data.frame(model),aes(x=LDH.preLD , y=yhat, z = Bulky)) + 
  geom_line(
    aes(col = Bulky)
  ) + 
  geom_ribbon(data = model, aes(ymin=lower, ymax=upper), alpha=0.2, linetype=0) + 
  theme_bw() + 
  labs(
    #title = "Survival by CAR-T product and ALC pre-LD",
     x = "LDH pre-LD (U/L)",
     y = "PFS (hazard)")+
  theme(title = element_text(size=16,face="bold"),legend.title=element_blank())

Hazard of OS: LDH - stratified by bulky status

Show code

data1 <- filter(dataset, Infused == 1) %>%
  mutate(
    Bulky = recode(Bulky, "1"="Bulky","0"="Not bulky")
  )

dd <- datadist(data1)
options(datadist='dd')

S <- Surv(data1$Days.to.DLC, data1$Death)
f <- cph(S ~ LDH.preLD * Bulky , x=TRUE, y=TRUE,surv=TRUE,data=data1)

med <- Quantile(f)
model <- Predict(f, Bulky, LDH.preLD, fun=function(x) exp(x) )

ggplot(as.data.frame(model),aes(x=LDH.preLD , y=yhat, z = Bulky)) + 
  geom_line(
    aes(col = Bulky)
  ) + 
  geom_ribbon(data = model, aes(ymin=lower, ymax=upper), alpha=0.2, linetype=0) + 
  theme_bw() + 
  labs(
    #title = "Survival by CAR-T product and ALC pre-LD",
     x = "LDH pre-LD (U/L)",
     y = "OS (hazard)")+
  theme(title = element_text(size=16,face="bold"),legend.title=element_blank())

Hazard of OS: LDH, stratified

Show code

data1 <- filter(dataset, Infused == 1)

dd <- datadist(data1)
options(datadist='dd')

S <- Surv(data1$Days.to.DLC, data1$Death)
f <- cph(S ~ LDH.preLD , x=TRUE, y=TRUE,surv=TRUE,data=data1)

med <- Quantile(f)
model <- Predict(f, LDH.preLD, fun=function(x) exp(x) )

ggplot(as.data.frame(model),aes(x=LDH.preLD , y=yhat)) + 
  geom_line() + 
  geom_ribbon(data = model, aes(ymin=lower, ymax=upper), alpha=0.2, linetype=0) + 
  theme_bw() + 
  labs(
    #title = "Survival by LDH pre-LD",
    x = "LDH pre-LD (U/L)",
    y = "OS (hazard)")+
  theme(title = element_text(size=16,face="bold"),legend.title=element_blank())

Hazard of PFS: LDH, stratified by product

Show code

data1 <- filter(dataset, Infused == 1) 

dd <- datadist(data1)
options(datadist='dd')

S <- Surv(data1$Days.to.relapse.death.or.DLC, data1$Relapse.or.death)
f <- cph(S ~ LDH.preLD * CAR.T.Product.Type , x=TRUE, y=TRUE,surv=TRUE,data=data1)



model <- Predict(f, CAR.T.Product.Type, LDH.preLD, fun=function(x) exp(x) )

ggplot(as.data.frame(model),aes(x=LDH.preLD, y=yhat, z = CAR.T.Product.Type)) + 
  geom_line(
    aes(col = CAR.T.Product.Type)
  ) + 
  geom_ribbon(data = model, aes(ymin=lower, ymax=upper), alpha=0.2, linetype=0) + 
  theme_bw() + 
  labs(
    #title = "Survival by product and LDH pre-LD",
    x = "LDH pre-LD (U/L)",
    y = "PFS (hazard)") +
  theme(title = element_text(size=16,face="bold"),legend.title=element_blank())

Hazard of PFS: LDH, stratified by bulky status

Show code

data1 <- filter(dataset, Infused == 1) %>%
  mutate(
    Bulky = recode(Bulky,"1"="Bulky","0"="Not bulky")
  )

dd <- datadist(data1)
options(datadist='dd')

S <- Surv(data1$Days.to.relapse.death.or.DLC, data1$Relapse.or.death)
f <- cph(S ~ LDH.preLD * Bulky , x=TRUE, y=TRUE,surv=TRUE,data=data1)

model <- Predict(f, Bulky, LDH.preLD, fun=function(x) exp(x) )

ggplot(as.data.frame(model),aes(x=LDH.preLD, y=yhat, z = Bulky)) + 
  geom_line(
    aes(col = Bulky)
  ) + 
  geom_ribbon(data = model, aes(ymin=lower, ymax=upper), alpha=0.2, linetype=0) + 
  theme_bw() + 
  labs(
    #title = "Survival by product and LDH pre-LD",
    x = "LDH pre-LD (U/L)",
    y = "PFS (hazard)") +
  theme(title = element_text(size=16,face="bold"),legend.title=element_blank())

Hazard of OS: LDH, stratified by bulky status

Show code

data1 <- filter(dataset, Infused == 1)%>%
  mutate(
    Bulky = recode(Bulky,"1"="Bulky","0"="Not bulky")
  )

dd <- datadist(data1)
options(datadist='dd')

S <- Surv(data1$Days.to.DLC, data1$Death)
f <- cph(S ~ LDH.preLD * Bulky , x=TRUE, y=TRUE,surv=TRUE,data=data1)

model <- Predict(f, Bulky, LDH.preLD, fun=function(x) exp(x) )

ggplot(as.data.frame(model),aes(x=LDH.preLD, y=yhat, z = Bulky)) + 
  geom_line(
    aes(col = Bulky)
  ) + 
  geom_ribbon(data = model, aes(ymin=lower, ymax=upper), alpha=0.2, linetype=0) + 
  theme_bw() + 
  labs(
    #title = "Survival by product and LDH pre-LD",
    x = "LDH pre-LD (U/L)",
    y = "OS (hazard)") +
  theme(title = element_text(size=16,face="bold"),legend.title=element_blank())

Hazard of PFS: LDH

Show code

data1 <- filter(dataset, Infused == 1)

dd <- datadist(data1)
options(datadist='dd')

S <- Surv(data1$Days.to.relapse.death.or.DLC, data1$Relapse.or.death)
f <- cph(S ~ LDH.preLD , x=TRUE, y=TRUE,surv=TRUE,data=data1)

model <- Predict(f, LDH.preLD, fun=function(x) exp(x) )

ggplot(as.data.frame(model),aes(x=LDH.preLD, y=yhat)) + 
  geom_line(
    aes()
  ) + 
  geom_ribbon(data = model, aes(ymin=lower, ymax=upper), alpha=0.2, linetype=0) + 
  theme_bw() + 
  labs(
    #title = "Survival by product and LDH pre-LD",
    x = "LDH pre-LD (U/L)",
    y = "PFS (hazard)") +
  theme(title = element_text(size=16,face="bold"),legend.title=element_blank())

Hazard of OS: Largest lesion

Show code

data1 <- dataset

dd <- datadist(data1)
options(datadist='dd')

S <- Surv(data1$Days.to.DLC, data1$Death)
f <- cph(S ~ Largest.lesion, x=TRUE, y=TRUE,surv=TRUE,data=data1)

med <- Quantile(f)
model <- Predict(f, Largest.lesion, fun=function(x) exp(x) )

ggplot(as.data.frame(model),aes(x=Largest.lesion , y=yhat)) + 
  geom_line() + 
  geom_ribbon(data = model, aes(ymin=lower, ymax=upper), alpha=0.2, linetype=0) + 
  theme_bw() + 
  labs(
    #title = "Survival by LDH pre-LD",
    x = "Largest lesion (cm)",
    y = "OS (hazard)")+
  theme(title = element_text(size=16,face="bold"),legend.title=element_blank())

Hazard of OS: Cr

Show code

data1 <- filter(dataset, Infused == 1) 

dd <- datadist(data1)
options(datadist='dd')

S <- Surv(data1$Days.to.DLC, data1$Death)
f <- cph(S ~ rcs(cre.preLD,3) , x=TRUE, y=TRUE,surv=TRUE,data=data1)



model <- Predict(f, cre.preLD, fun=function(x) exp(x) )

ggplot(as.data.frame(model),aes(x=cre.preLD, y=yhat)) + 
  geom_line(
    aes()
  ) + 
  geom_ribbon(data = model, aes(ymin=lower, ymax=upper), alpha=0.2, linetype=0) + 
  theme_bw() + 
  labs(
    #title = "Survival by product and pre-LD Cr",
    x = "Pre-LD Cr (U/L)",
    y = "OS (hazard)") +
  theme(title = element_text(size=16,face="bold"),legend.title=element_blank())

CURVE SMOOTHING

ANC Loess modeling – stratified by product

Show code

data1 <- CBCdata %>%
  mutate(
    CAR.T.Product.Type = recode(CAR.T.Product.Type, "Breyanzi" = "Liso-cel", "Yescarta" = "Axi-cel")
  )

ggplot() + 
  geom_smooth(data = filter(data1, CAR.T.Product.Type == "Liso-cel"), aes(x = day, y = ancx, color = "Liso-cel"), method = "loess", formula = y~x, fill = "#69b3a2", se = TRUE, na.rm = TRUE) + 
  geom_smooth(data = filter(data1, CAR.T.Product.Type == "Axi-cel"), aes(x = day, y = ancx, color = "Axi-cel"), method = "loess", formula = y~x, fill = "#69b3a2", se = TRUE, na.rm = TRUE) + 
  labs(
    x = "Time relative to infusion (days)",
    y = "Absolute neutrophil count (cells/µL)",
    color = "Legend"
  ) + 
  scale_y_continuous(breaks = c(0, 1000, 2000, 3000, 4000, 5000, 6000)) +
  xlim(-5, 50) +
  theme_classic() + 
  theme(legend.position = c(0.8, 0.9)) +
  scale_color_manual(values = c("steelblue", "coral2")) +
  theme(text = element_text(size = 16))

Plt Loess modeling – stratified by product

Show code

data1 <- CBCdata %>%
  mutate(
    CAR.T.Product.Type = recode(CAR.T.Product.Type, "Breyanzi" = "Liso-cel", "Yescarta" = "Axi-cel")
  )

ggplot() + 
  geom_smooth(data = filter(data1, CAR.T.Product.Type == "Liso-cel"), aes(x=day, y=TotPlateletCnt/1000, color = "Liso-cel"), method="loess" , formula = y~x, fill="#69b3a2", se=TRUE, na.rm = TRUE) + 
  geom_smooth(data = filter(data1, CAR.T.Product.Type == "Axi-cel"), aes(x=day, y=TotPlateletCnt/1000, color = "Axi-cel"),method="loess" , formula = y~x , fill="#69b3a2", se=TRUE, na.rm = TRUE) + 
  labs(x = "Time relative to infusion (days)",
       y = "Platelet count (10\u00B3/\u00B5L)",
       color = "Legend") + 
  theme_classic() + 
  xlim(-5,50) +
  scale_color_manual(values = c ("steelblue","coral2")) + 
  theme(legend.position=c(.8,.9)) +
  theme(text = element_text(size = 16))  

Ferritin Loess modeling – stratified by CRS

Show code

ferritin.data = filter(labs, day >=-5 & day <=30) %>%
  select(day, time, ferritin_ng_ml, uwid) %>%
  mutate(
    time.percent = sapply( strsplit(time,":"),
                           function(x) {
                             x <- as.numeric(x)
                             x[1]/24
                           }
    ),
    day = ifelse(day>=0, day + time.percent, day - time.percent),
    ferritin = log10(ferritin_ng_ml)
  )

ferritin.data <-
  left_join(
    x = ferritin.data,
    y = dataset %>%
      select(CAR.T.Product.Type, CRS.grade, uwid, Study.No),
    by = "uwid"
  )

ggplot() + 
    geom_smooth(data = filter(ferritin.data, CRS.grade == "3" | CRS.grade == "4"), aes(x=day, y=ferritin_ng_ml, color = "G\u22653 CRS"),method="loess" , formula = y~x , fill="#69b3a2", se=TRUE, na.rm = TRUE) + 
  geom_smooth(data = filter(ferritin.data, CRS.grade == "1" | CRS.grade == "2"), aes(x=day, y=ferritin_ng_ml, color = "G1-2 CRS"),method="loess" , formula = y~x , fill="#69b3a2", se=TRUE, na.rm = TRUE) + 
    geom_smooth(data = filter(ferritin.data, CRS.grade == "0"), aes(x=day, y=ferritin_ng_ml, color = "G0 CRS"),method="loess" , formula = y~x , fill="#69b3a2", se=TRUE, na.rm = TRUE) + 
  labs(x = "Time relative to infusion (days)",
       y = "Ferritin (ng/mL)",
       color = "Legend") +
  theme_classic() + 
  theme(legend.position=c(.8,.8)) +
  theme(text = element_text(size = 16))  +
  scale_color_manual(values = c ("coral2","purple","steelblue")) +
  coord_cartesian(ylim = c(0, 9000), xlim = c(0, 30))

Ferritin Loess modeling – stratified by ICANS

Show code

ferritin.data = filter(labs, day >=-5 & day <=30) %>%
  select(day, time, ferritin_ng_ml, uwid) %>%
  mutate(
    time.percent = sapply( strsplit(time,":"),
                           function(x) {
                             x <- as.numeric(x)
                             x[1]/24
                           }
    ),
    day = ifelse(day>=0, day + time.percent, day - time.percent),
    ferritin = log10(ferritin_ng_ml)
  )

ferritin.data <-
  left_join(
    x = ferritin.data,
    y = dataset %>%
      select(CAR.T.Product.Type, CRS.grade, ICANS.grade, uwid, Study.No),
    by = "uwid"
  )

ggplot() + 
    geom_smooth(data = filter(ferritin.data, ICANS.grade == "3" | CRS.grade == "4"), aes(x=day, y=ferritin_ng_ml, color = "G\u22653 ICANS"),method="loess" , formula = y~x , fill="#69b3a2", se=TRUE, na.rm = TRUE) + 
  geom_smooth(data = filter(ferritin.data, ICANS.grade == "1" | CRS.grade == "2"), aes(x=day, y=ferritin_ng_ml, color = "G1-2 ICANS"),method="loess" , formula = y~x , fill="#69b3a2", se=TRUE, na.rm = TRUE) + 
    geom_smooth(data = filter(ferritin.data, ICANS.grade == "0"), aes(x=day, y=ferritin_ng_ml, color = "G0 ICANS"),method="loess" , formula = y~x , fill="#69b3a2", se=TRUE, na.rm = TRUE) + 
  labs(x = "Time relative to infusion (days)",
       y = "Ferritin (ng/mL)",
       color = "Legend") + 
  theme_classic() + 
  theme(legend.position=c(.2,.8)) +
  theme(text = element_text(size = 16))  +
  scale_color_manual(values = c ("coral2","purple","steelblue")) +
  coord_cartesian(ylim = c(0, 9000), xlim = c(0, 30))

Ferritin Loess modeling – stratified by response

Show code

ferritin.data = filter(labs, day >=-5 & day <=30) %>%
  select(day, time, ferritin_ng_ml, uwid) %>%
  mutate(
    time.percent = sapply( strsplit(time,":"),
                           function(x) {
                             x <- as.numeric(x)
                             x[1]/24
                           }
    ),
    day = ifelse(day>=0, day + time.percent, day - time.percent),
    ferritin = log10(ferritin_ng_ml)
  )

ferritin.data <-
  left_join(
    x = ferritin.data,
    y = dataset %>%
      select(CAR.T.Product.Type, CRS.grade, ICANS.grade, uwid, Study.No, Best.CR, Best.PR, Best.PD),
    by = "uwid"
  )

ggplot() + 
    geom_smooth(data = filter(ferritin.data, Best.CR == 1), aes(x=day, y=ferritin_ng_ml, color = "CR"),method="loess" , formula = y~x , fill="#69b3a2", se=TRUE, na.rm = TRUE) + 
  geom_smooth(data = filter(ferritin.data, Best.PR == 1), aes(x=day, y=ferritin_ng_ml, color = "PR"),method="loess" , formula = y~x , fill="#69b3a2", se=TRUE, na.rm = TRUE) + 
    geom_smooth(data = filter(ferritin.data, Best.PD == 1), aes(x=day, y=ferritin_ng_ml, color = "PD"),method="loess" , formula = y~x , fill="#69b3a2", se=TRUE, na.rm = TRUE) + 
  labs(x = "Time relative to infusion (days)",
       y = "Ferritin (ng/mL)",
       color = "Legend") + 
  theme_classic() + 
  theme(legend.position=c(.2,.8)) +
  theme(text = element_text(size = 16))  +
  scale_color_manual(values = c ("coral2","purple","steelblue")) +
  coord_cartesian(ylim = c(0, 9000), xlim = c(0, 30))

Ferritin Loess modeling – stratified by product

Show code

ferritin.data = filter(labs, day >=-5 & day <=30) %>%
  select(day, time, ferritin_ng_ml, uwid) %>%
  mutate(
    time.percent = sapply( strsplit(time,":"),
                           function(x) {
                             x <- as.numeric(x)
                             x[1]/24
                           }
    ),
    day = ifelse(day>=0, day + time.percent, day - time.percent),
    ferritin = log10(ferritin_ng_ml)
  )

ferritin.data <-
  left_join(
    x = ferritin.data,
    y = dataset %>%
      select(CAR.T.Product.Type, CRS.grade, ICANS.grade, uwid, Study.No, Best.CR, Best.PR, Best.PD),
    by = "uwid"
  )

ggplot() + 
    geom_smooth(data = filter(ferritin.data, CAR.T.Product.Type == "Breyanzi"), aes(x=day, y=ferritin_ng_ml, color = "Liso-cel"),method="loess" , formula = y~x , fill="#69b3a2", se=TRUE, na.rm = TRUE) + 
  geom_smooth(data = filter(ferritin.data, CAR.T.Product.Type == "Yescarta"), aes(x=day, y=ferritin_ng_ml, color = "Axi-cel"),method="loess" , formula = y~x , fill="#69b3a2", se=TRUE, na.rm = TRUE) + 
  labs(x = "Time relative to infusion (days)",
       y = "Ferritin (ng/mL)",
       color = "Legend") + 
  theme_classic() + 
  theme(legend.position=c(.2,.8)) +
  theme(text = element_text(size = 16))  +
  scale_color_manual(values = c ("coral2","purple","steelblue")) +
  coord_cartesian(ylim = c(0, 5000), xlim = c(0, 30))

CONTOUR PLOTS

Contour plot for largest lesion and LDH pre-LD

Show code

data1 <- filter(dataset, CAR.T.Product.Type == "Breyanzi")
dd <- datadist(data1)
options(datadist='dd')

S <- Surv(data1$Days.to.DLC/30, data1$Death)
f <- cph(S ~ Largest.lesion * LDH.preLD , x=TRUE, y=TRUE,surv=TRUE,data=data1)


med <- Quantile(f)
data.frame(Predict(f,Largest.lesion,LDH.preLD,fun=function(x) med(lp=x))) %>%
  ggplot() +
  geom_contour_filled(aes(Largest.lesion,LDH.preLD, z = yhat)) +
  theme_classic() +
  labs(title = "Liso-cel",
       x = "Largest lesion (cm)",
       y = "LDH pre-LD (U/L)",
       fill="Median survival (months)")  +
  coord_cartesian(xlim = c(0, 10), ylim = c(100, 700))

Show code

data1 <- filter(dataset, CAR.T.Product.Type == "Yescarta") %>%
  select(Largest.lesion, LDH.preLD, Days.to.DLC, Death
  )
dd <- datadist(data1)
options(datadist='dd')

S <- Surv(data1$Days.to.DLC/30, data1$Death)
f <- cph(S ~ Largest.lesion * LDH.preLD , x=TRUE, y=TRUE,surv=TRUE,data=data1)

med <- Quantile(f)
data.frame(Predict(f,Largest.lesion,LDH.preLD,fun=function(x) med(lp=x))) %>%
  ggplot() +
  geom_contour_filled(aes(Largest.lesion,LDH.preLD, z = yhat)) +
  theme_classic() +
  labs(title = "Axi-cel",
       x = "Largest lesion (cm)",
       y = "LDH pre-LD (U/L)",
       fill="Median survival (months)") +
  coord_cartesian(xlim = c(0, 10), ylim = c(100, 700))

Contour plot for Age and ALC pre-LD

Show code

data1 <- filter(dataset, Infused == 1)
dd <- datadist(data1)
options(datadist='dd')

S <- Surv(data1$Days.to.DLC, data1$Death)
f <- cph(S ~ Age * ALC.preLD , x=TRUE, y=TRUE,surv=TRUE,data=data1)

med <- Quantile(f)
data.frame(Predict(f,Age,ALC.preLD,fun=function(x) med(lp=x)/365.25)) %>%
  ggplot() +
  geom_contour_filled(aes(Age,ALC.preLD, z = yhat)) +
  theme_classic() +
  labs(title = "Survival: ALC pre-LD and Age",
       x = "Age (years)",
       y = "ALC pre-LD (10\U00B3/\U00B5L)",
       fill="Median survival (years)") 

Swimmer’s plot

Swimmer plot

Show code

swimmer_data = as.data.frame(read_excel(path = "Swimmer.xlsx")) %>%
  mutate(
    Response_start = Response_start/30,
    Response_end = Response_end/30,
    Time = Time/30,
    Product = recode(Product, "Breyanzi" = "Liso-cel","Yescarta"="Axi-cel"),
    Response = factor(Response, levels = c("CR","PR","SD", "PD"))
  )

pastel1_palette <- brewer.pal(4, "Pastel1")[c(3,2,4,1)]


swimmer <- swimmer_plot(df=filter(swimmer_data, !is.na(Response) ), 
            id="Number",
            end= "Response_end", 
            name_fill="Response",
            stratify = "Product",
            ) +
  scale_fill_manual(values = pastel1_palette) +
  swimmer_arrows(df_arrows = filter(swimmer_data,!is.na(Continued_response) ),
                 id="Number",
                 arrow_start="Response_end", 
                 name_col = 'Response', 
                 type = "closed", cex=.5, 
                 arrow_positions = c(1,6)
                 )+ 
  scale_color_manual(name="Response",values=pastel1_palette,drop=FALSE) +
  swimmer_points(df_points= filter(swimmer_data,!is.na(Event) ) ,id='Number',name_shape = "Event", time='Time',size=1,fill='white',col="black") +
  scale_y_continuous(name = "Time since infusion (Months)",breaks = seq(0,10000,by=12)) +
  guides(fill = guide_legend(title = "Response"), colour = FALSE)  # Hide the legend for arrows
swimmer